popnn::rnn::RnnParams Struct Reference

Structure of Recurrent Neural Network (RNN) parameters which allows for any customized implementation of the cellular part of the RNN. More...

#include <Rnn.hpp>

Public Attributes
poplar::Type	dataType
	The datatype used for the RNN.
std::size_t	batchSize
	The batch size.
std::size_t	maxTimeSteps
	The maximum number of RNN time steps.
std::size_t	timeSteps
poplar::Tensor	varTimeSteps
	The run-time number of RNN time steps of dimension [batchSize] If this tensor is default constructed, the number of time steps for the sequence corresponding to each batch will be set according to the maxTimeSteps member.
std::vector< std::size_t >	layerSizes
	For each RNN layer, the layer size parameter needs to be specified for the input and the output. More...

Detailed Description

Structure of Recurrent Neural Network (RNN) parameters which allows for any customized implementation of the cellular part of the RNN.

Member Data Documentation

layerSizes

std::vector<std::size_t> popnn::rnn::RnnParams::layerSizes

For each RNN layer, the layer size parameter needs to be specified for the input and the output.

This is done using a 2-element vector where the first element is the input size and the second element is the output size of the RNN layer.

timeSteps

std::size_t popnn::rnn::RnnParams::timeSteps

Deprecated:

Use RnnParams.maxTimeSteps instead.

---

The documentation for this struct was generated from the following file:

• include/popnn/Rnn.hpp