Core types and functions

Under construction

In the meantime, please refer to the API Reference of deepali.core.

Common types and functions that operate on tensors representing different types of data.

Besides defining common types and auxiliary functions extending the standard library, this core library in particular defines functions which operate on objects of type torch.Tensor. This functional API defines a set of reusable state-less functions similar to torch.nn.functional. Object-oriented APIs use this functional API to realize their functionality. In particular, the forward() method of torch.nn.Module subclasses, such as for example data transformations (cf. data.transforms) and neural network components (cf. modules and networks) are implemented using these functional building blocks.

The following import statement can be used to access the functional API:

import deepali.core.functional as U