`torch_adata._tools`¶

Submodules¶

Package Contents¶

Classes¶

`BaseLightningDataModule`	Helper class that provides a standard way to create an ABC using
`AnnDatasetSplit`

Functions¶

`split`(dataset, n_groups, percentages, , ...)	Split dataset using torch.utils.data.random_split.
`idx`(→ numpy.ndarray)	Create a an index, sampling from range(len(dataset)).
`dummy_batch`(dataset[, silent])	Fetch a dummy batch

Attributes¶

`__module_name__`
`__doc__`
`__author__`
`__email__`
`__version__`

torch_adata._tools.__module_name__ = __init__.py¶

torch_adata._tools.__doc__ = functions __init__ module.¶

torch_adata._tools.__author__¶

torch_adata._tools.__email__¶

torch_adata._tools.__version__ = 0.0.20¶

torch_adata._tools.split(dataset: torch.utils.data.Dataset, n_groups: int = 2, percentages: list([float, '...', float]) = None)¶

Split dataset using torch.utils.data.random_split.

dataset: type: torch.utils.data.Dataset
n_groups: type: int default: 2
percentages: type: list([float, …, float]) default: None

list([torch.utils.data.Dataset, …, torch.utils.data.Dataset])

Uses the torch.utils.data.random_split function to actually do the split.

torch_adata._tools.idx(dataset, size: int = None, replace: bool = False) → numpy.ndarray¶

Create a an index, sampling from range(len(dataset)).

dataset

size: type: int
replace: type: bool

idx: type: numpy.ndarray

torch_adata._tools.dummy_batch(dataset, silent=False)¶: Fetch a dummy batch

class torch_adata._tools.BaseLightningDataModule(adata: anndata.AnnData = None, batch_size: int = 2000, num_workers: int = os.cpu_count(), **kwargs)¶

Bases: abc.ABC, pytorch_lightning.LightningDataModule

Helper class that provides a standard way to create an ABC using inheritance.

__parse__(kwargs, ignore=['self', '__class__'])¶

train_dataloader()¶

val_dataloader()¶

test_dataloader()¶

predict_dataloader()¶

class torch_adata._tools.AnnDatasetSplit(adata: anndata.AnnData, use_key: str = 'X_pca', groupby: str = None, obs_keys: str = None, percent_val: float = 0.2, train_key: str = 'train', test_key: str = 'test', **kwargs)¶

__parse__(kwargs, ignore=['self'])¶

_train_test_obs_keys()¶

to_dataset(adata)¶

on_test_train()¶: Split cells in adata on train / test columns in adata.obs.

allocate_validation()¶: Split the previously allocated train_dataset into training and validation subsets.