dgs.utils.types

definition of regularly used types

Module Attributes

Config	A nested configuration, describing a module or the whole tracker.
NodePath	A list of key-names used for traversing through a config.
FilePath	The path to a file or directory.
FilePaths	Multiple FilePaths as tuple.
Validator	A function for validating a value of a Config object.
Validations	A dictionary of validations, mapping a value of a given Config to some sort of validation.
DataGetter	Function to extract a single specific value from a State.
DataGetters	Function to extract multiple attributes or properties from a State.
Instance	An instance to be loaded is either the name of that instance or a class-type.
Metric	A module or function that computes a metric.
Loss	A module or function that computes a loss.
Scheduler	A Scheduler used to update the learning rate during training.
Results	A dictionary of results, as name->metric or name->iterable of metrics.
Device	A Torch device, either descriptive string (e.g. 'cpu') or a regular torch.device object.
Video	A tensor based video, either torchvision or regular tensor.
Image	A tensor based image with shape [B x C x H x W].
Images	A list with length B, containing Images with shape [1 x C x H x W].
Heatmap	Heatmap as mask with shape [(B x) J x h x w].
ImgShape	Shape of an image as tuple like (height, width).