TensorTrait Constructor

Product Support

Product	Supported
Atlas A3 training products/Atlas A3 inference products	√
Atlas A2 training products/Atlas A2 inference products	√
Atlas 200I/500 A2 inference products	x
Atlas inference product's AI Core	x
Atlas inference product's Vector Core	x
Atlas training products	x

Function

Instantiates a TensorTrait object based on the input Layout object.

Prototype

__aicore__ inline TensorTrait(const LayoutType& t = {})

Parameters

Parameter	Input/Output	Description
t	Input	Input Layout object. The input data type LayoutType must meet the requirements described in Restrictions.

Returns

None

Restrictions

None

Example

TensorTrait usage example

// The following is a unit test example of using TensorTrait based on GoogleTest.

// Use the MakeTensorTrait method to create a TensorTrait object.
AscendC::Shape<int,int,int> shape = AscendC::MakeShape(10, 20, 30);
AscendC::Stride<int,int,int> stride = AscendC::MakeStride(1, 100, 200);
auto layoutMake = AscendC::MakeLayout(shape, stride);    
auto tensorTraitMake = AscendC::MakeTensorTrait<float, AscendC::TPosition::VECIN>(layoutMake);

EXPECT_EQ(AscendC::Std::get<0>(tensorTraitMake.GetLayout().GetShape()), 10);
EXPECT_EQ(AscendC::Std::get<1>(tensorTraitMake.GetLayout().GetShape()), 20);
EXPECT_EQ(AscendC::Std::get<2>(tensorTraitMake.GetLayout().GetShape()), 30);
EXPECT_EQ(AscendC::Std::get<0>(tensorTraitMake.GetLayout().GetStride()), 1);
EXPECT_EQ(AscendC::Std::get<1>(tensorTraitMake.GetLayout().GetStride()), 100);
EXPECT_EQ(AscendC::Std::get<2>(tensorTraitMake.GetLayout().GetStride()), 200);

// Use the constructor to create a TensorTrait object.
using TensorTraitType = AscendC::TensorTrait<half, AscendC::TPosition::VECCALC, AscendC::Layout<AscendC::Shape<int, int, int>, AscendC::Stride<int, int, int>>>;
TensorTraitType tensorTraitInit(layoutMake);

EXPECT_EQ(AscendC::Std::get<0>(tensorTraitInit.GetLayout().GetShape()), 10);
EXPECT_EQ(AscendC::Std::get<1>(tensorTraitInit.GetLayout().GetShape()), 20);
EXPECT_EQ(AscendC::Std::get<2>(tensorTraitInit.GetLayout().GetShape()), 30);
EXPECT_EQ(AscendC::Std::get<0>(tensorTraitInit.GetLayout().GetStride()), 1);
EXPECT_EQ(AscendC::Std::get<1>(tensorTraitInit.GetLayout().GetStride()), 100);
EXPECT_EQ(AscendC::Std::get<2>(tensorTraitInit.GetLayout().GetStride()), 200);

EXPECT_EQ(AscendC::Std::get<0>(tensorTraitInit.GetShape()), 10);
EXPECT_EQ(AscendC::Std::get<1>(tensorTraitInit.GetShape()), 20);
EXPECT_EQ(AscendC::Std::get<2>(tensorTraitInit.GetShape()), 30);
EXPECT_EQ(AscendC::Std::get<0>(tensorTraitInit.GetStride()), 1);
EXPECT_EQ(AscendC::Std::get<1>(tensorTraitInit.GetStride()), 100);
EXPECT_EQ(AscendC::Std::get<2>(tensorTraitInit.GetStride()), 200);

// Use the SetLayout method to set TensorTrait.
TensorTraitType tensorTraitSet;
tensorTraitSet.SetLayout(layoutMake);

EXPECT_EQ(AscendC::Std::get<0>(tensorTraitSet.GetLayout().GetShape()), 10);
EXPECT_EQ(AscendC::Std::get<1>(tensorTraitSet.GetLayout().GetShape()), 20);
EXPECT_EQ(AscendC::Std::get<2>(tensorTraitSet.GetLayout().GetShape()), 30);
EXPECT_EQ(AscendC::Std::get<0>(tensorTraitSet.GetLayout().GetStride()), 1);
EXPECT_EQ(AscendC::Std::get<1>(tensorTraitSet.GetLayout().GetStride()), 100);
EXPECT_EQ(AscendC::Std::get<2>(tensorTraitSet.GetLayout().GetStride()), 200);

TensorTrait and API usage example

AscendC::LocalTensor<AscendC::TensorTrait<half>> tensor1 = que1.DeQue<AscendC::TensorTrait<half>>();
AscendC::LocalTensor<AscendC::TensorTrait<half>> tensor2 = que2.DeQue<AscendC::TensorTrait<half>>();
AscendC::LocalTensor<AscendC::TensorTrait<half>> tensor3 = que3.AllocTensor<AscendC::TensorTrait<half>>();
Add(tensor3, tensor1, tensor2, tensor3.GetSize());

Parent topic: TensorTrait