The family's state shape: one layer per family (a single scalar TCN layer, one LSTM/GRU layer, one diagonal SSM). More...

#include <gpu_neural_model.hpp>

Public Attributes
std::uint32_t	tcn_kernel_size = 0
	Tcn: one scalar dilated causal layer.
std::uint32_t	tcn_dilation = 0
std::uint32_t	recurrent_hidden_size = 0
	Lstm, Gru: one scalar-input layer.
std::uint32_t	ssm_state_size = 0
	Ssm: a diagonal state space.

Detailed Description

The family's state shape: one layer per family (a single scalar TCN layer, one LSTM/GRU layer, one diagonal SSM).

Deeper models fit through a kernel that owns its topology. Only the fields for the requested family are read; WaveNet, Custom and "nam.tcn" kernels own their topology.

Member Data Documentation

◆ recurrent_hidden_size

std::uint32_t pulp::gpu_audio::GpuNeuralModelShape::recurrent_hidden_size = 0

Lstm, Gru: one scalar-input layer.

◆ ssm_state_size

std::uint32_t pulp::gpu_audio::GpuNeuralModelShape::ssm_state_size = 0

Ssm: a diagonal state space.

◆ tcn_dilation

std::uint32_t pulp::gpu_audio::GpuNeuralModelShape::tcn_dilation = 0

◆ tcn_kernel_size

std::uint32_t pulp::gpu_audio::GpuNeuralModelShape::tcn_kernel_size = 0

Tcn: one scalar dilated causal layer.

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

/home/runner/work/pulp/pulp/core/gpu_audio/include/pulp/gpu_audio/gpu_neural_model.hpp