pulp::gpu_audio::GpuNeuralModelRequest Struct Reference

#include <gpu_neural_model.hpp>

Public Attributes
NeuralFamily	family = NeuralFamily::Unknown
GpuNeuralModelManifest	manifest {}
GpuNeuralModelShape	shape {}
std::uint32_t	block_size = 0
std::span< const GpuNeuralCpuKernel >	channel_kernels {}
	One kernel per audio channel.
const GpuWaveNetSession::Config *	wavenet_gpu = nullptr
	WaveNet only: run the GPU path with these session settings (copied).
std::uint32_t	gpu_max_inflight = 0
	GPU blocks in flight; at most the session's slots.

Member Data Documentation

block_size

std::uint32_t pulp::gpu_audio::GpuNeuralModelRequest::block_size = 0

channel_kernels

std::span<const GpuNeuralCpuKernel> pulp::gpu_audio::GpuNeuralModelRequest::channel_kernels {}

One kernel per audio channel.

family

NeuralFamily pulp::gpu_audio::GpuNeuralModelRequest::family = NeuralFamily::Unknown

gpu_max_inflight

std::uint32_t pulp::gpu_audio::GpuNeuralModelRequest::gpu_max_inflight = 0

GPU blocks in flight; at most the session's slots.

A node using this model may not set a larger max_inflight. Zero selects two when the session has at least two slots, else one.

manifest

GpuNeuralModelManifest pulp::gpu_audio::GpuNeuralModelRequest::manifest {}

shape

GpuNeuralModelShape pulp::gpu_audio::GpuNeuralModelRequest::shape {}

wavenet_gpu

const GpuWaveNetSession::Config* pulp::gpu_audio::GpuNeuralModelRequest::wavenet_gpu = nullptr

WaveNet only: run the GPU path with these session settings (copied).

Null builds a CPU-only model. A provider that cannot be prepared leaves the node on its CPU model, reported Degraded.

---

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