mirror of
https://github.com/saymrwulf/onnxruntime.git
synced 2026-05-29 23:06:41 +00:00
* Implement conversion from ortvalue to Itensor for string tensors and comparing sequence of maps of strings to floats * PR comments
84 lines
No EOL
2.8 KiB
C++
84 lines
No EOL
2.8 KiB
C++
#include "testPch.h"
|
|
#include "onnxruntime_cxx_api.h"
|
|
|
|
namespace OrtValueHelpers {
|
|
winml::ITensor LoadTensorFromOrtValue(Ort::Value& val);
|
|
|
|
Ort::Value CreateOrtValueFromITensor(winml::ITensor winmlTensor);
|
|
}
|
|
|
|
template <ONNXTensorElementDataType T>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind {
|
|
// Invalid ONNXTensorElementDataType to TensorKind
|
|
static_assert(sizeof(T) == -1, "No WinML TensorKind mapped for given ONNX Tensor Element type!");
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT> {
|
|
typedef winml::TensorFloat Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT8> {
|
|
typedef winml::TensorUInt8Bit Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_INT8> {
|
|
typedef winml::TensorInt8Bit Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT16> {
|
|
typedef winml::TensorUInt16Bit Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_INT16> {
|
|
typedef winml::TensorInt16Bit Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_INT32> {
|
|
typedef winml::TensorInt32Bit Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_INT64> {
|
|
typedef winml::TensorInt64Bit Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_STRING> {
|
|
typedef winml::TensorString Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_BOOL> {
|
|
typedef winml::TensorBoolean Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT16> {
|
|
typedef winml::TensorFloat16Bit Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_DOUBLE> {
|
|
typedef winml::TensorDouble Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT32> {
|
|
typedef winml::TensorUInt32Bit Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_UINT64> {
|
|
typedef winml::TensorUInt64Bit Type;
|
|
};
|
|
|
|
template <>
|
|
struct ONNXTensorElementDataTypeToWinMLTensorKind<ONNXTensorElementDataType::ONNX_TENSOR_ELEMENT_DATA_TYPE_BFLOAT16> {
|
|
typedef winml::TensorFloat16Bit Type;
|
|
}; |