norse.torch.functional.decode module

Stateless decoding functionality for Norse, where events in time are converted to numerical representations. Many of the decoders are simply mirrors of the PyTorch Tensor aggregation functions, with the difference that they typically aggregate the first (temporal) dimension. They are therefore useful to inject in norse.torch.functional.lift() ed modules or situations where you need pickled functions.

norse.torch.functional.decode.sum_decode(tensor, dimension=0)[source]

Sums the input tensor in the first dimension by default.