eve::tensor::OnnxBuffer结构体 参考

Immutable buffer shared by graph aliases; exactly one storage is authoritative. 更多...

#include <OnnxStorage.h>

Public 属性
std::shared_ptr< const std::vector< uint8_t > >	host
std::shared_ptr< OnnxDeviceStorage >	device
size_t	size = 0
bool	persistent = false

详细描述

Immutable buffer shared by graph aliases; exactly one storage is authoritative.

注解

host owns CPU bytes; device owns GPU bytes. Host inputs may be cached by shared identity across calls only when persistent marks an immutable model initializer. Consumers use the device thread.

在文件 OnnxStorage.h 第 19 行定义.

类成员变量说明

device

std::shared_ptr<OnnxDeviceStorage> eve::tensor::OnnxBuffer::device

在文件 OnnxStorage.h 第 21 行定义.

host

std::shared_ptr<const std::vector<uint8_t> > eve::tensor::OnnxBuffer::host

在文件 OnnxStorage.h 第 20 行定义.

persistent

bool eve::tensor::OnnxBuffer::persistent = false

在文件 OnnxStorage.h 第 23 行定义.

size

size_t eve::tensor::OnnxBuffer::size = 0

在文件 OnnxStorage.h 第 22 行定义.

---

该结构体的文档由以下文件生成:

• src/modules/tensor/OnnxStorage.h