antimatter.handlers.pytorch_dataloader#

Module Contents#

Classes#

PytorchDataLoader

The PytorchDataLoader DataHandler supports a pytorch DataLoader. There are
class antimatter.handlers.pytorch_dataloader.PytorchDataLoader#

Bases: antimatter.handlers.base.DataHandler

The PytorchDataLoader DataHandler supports a pytorch DataLoader. There are some restrictions on the underlying dataset, which must be iterable, producing two-dimensional dictionaries.

from_generic(cols: List[str], generic_data: List[List[bytes]], extra: Dict[str, Any]) → Any#

from_generic loads the generic data as a dataset into the pytorch DataLoader, passing any extra parameters transparently to the DataLoader constructor

Parameters:

  • cols – the column names for the underlying data

  • generic_data – the capsule’s generic data format that is loaded into a pytorch DataLoader

  • extra – extra data for the DataHandler, passed into the pytorch DataLoader constructor

Returns:

the pytorch DataLoader built with the dataset

to_generic(dl: Any) → Tuple[List[str], List[List[bytes]], Dict[str, Any]]#

to_generic converts a pytorch DataLoader into the generic data format, iterating through the DataLoader’s data set, expecting each iterated item to be a 2-dimensional dictionary.

Parameters:

dl – the DataLoader to extract generic format data from

Returns:

the data in its generic form