How to initialize onnx input CreateTensor with gpu meory instead of CreateCpu

[wangzhenlin123]

Describe the issue

hi，How to initialize onnx input CreateTensor with gpu meory instead of CreateCpu？I haven't found a solution yet。

To reproduce

Ort::MemoryInfo memory_info = Ort::MemoryInfo::CreateCpu(OrtArenaAllocator, OrtMemTypeDefault);
//Ort::MemoryInfo memory_info = Ort::MemoryInfo::CreateCuda(OrtDeviceAllocator, 0);
Ort::Value input_tensor = Ort::Value::CreateTensor(memory_info, input_tensor_values.data(),
input_tensor_size, input_node_dims.data(), 4);

Urgency

No response

Platform

Windows

OS Version

win10

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

1.8.1 or 1.16

ONNX Runtime API

C++

Architecture

X64

Execution Provider

CUDA

Execution Provider Library Version

No response

[pranavsharma]

This API helps you create memory info for any EP.

onnxruntime/include/onnxruntime/core/session/onnxruntime_c_api.h

Lines 1554 to 1565 in b2aec41

	/** \brief Create an ::OrtMemoryInfo
	*
	* \param[in] name
	* \param[in] type
	* \param[in] id
	* \param[in] mem_type
	* \param[out] out Newly created ::OrtMemoryInfo. Must be freed with OrtAPi::ReleaseMemoryInfo
	*
	* \snippet{doc} snippets.dox OrtStatus Return Value
	*/
	ORT_API2_STATUS(CreateMemoryInfo, _In_ const char* name, enum OrtAllocatorType type, int id,
	enum OrtMemType mem_type, _Outptr_ OrtMemoryInfo** out);

[yuslepukhin]

auto cuda_mem_info = Ort::MemoryInfo{"cuda", OrtArenaAllocator, OrtMemTypeDefault};

[github-actions]

This issue has been automatically marked as stale due to inactivity and will be closed in 30 days if no further activity occurs. If further support is needed, please provide an update and/or more details.