Sdílet prostřednictvím


TensorInt64Bit Class

Definition

A 64-bit signed integer tensor object.

public ref class TensorInt64Bit sealed : ITensor
public ref class TensorInt64Bit sealed : ITensor, IClosable, IMemoryBuffer
/// [Windows.Foundation.Metadata.ContractVersion(Windows.AI.MachineLearning.MachineLearningContract, 65536)]
/// [Windows.Foundation.Metadata.MarshalingBehavior(Windows.Foundation.Metadata.MarshalingType.Agile)]
/// [Windows.Foundation.Metadata.Threading(Windows.Foundation.Metadata.ThreadingModel.Both)]
class TensorInt64Bit final : ITensor
/// [Windows.Foundation.Metadata.ContractVersion(Windows.AI.MachineLearning.MachineLearningContract, 65536)]
/// [Windows.Foundation.Metadata.MarshalingBehavior(Windows.Foundation.Metadata.MarshalingType.Agile)]
/// [Windows.Foundation.Metadata.Threading(Windows.Foundation.Metadata.ThreadingModel.Both)]
class TensorInt64Bit final : ITensor, IClosable, IMemoryBuffer
[Windows.Foundation.Metadata.ContractVersion(typeof(Windows.AI.MachineLearning.MachineLearningContract), 65536)]
[Windows.Foundation.Metadata.MarshalingBehavior(Windows.Foundation.Metadata.MarshalingType.Agile)]
[Windows.Foundation.Metadata.Threading(Windows.Foundation.Metadata.ThreadingModel.Both)]
public sealed class TensorInt64Bit : ITensor
[Windows.Foundation.Metadata.ContractVersion(typeof(Windows.AI.MachineLearning.MachineLearningContract), 65536)]
[Windows.Foundation.Metadata.MarshalingBehavior(Windows.Foundation.Metadata.MarshalingType.Agile)]
[Windows.Foundation.Metadata.Threading(Windows.Foundation.Metadata.ThreadingModel.Both)]
public sealed class TensorInt64Bit : System.IDisposable, ITensor, IMemoryBuffer
Public NotInheritable Class TensorInt64Bit
Implements ITensor
Public NotInheritable Class TensorInt64Bit
Implements IDisposable, IMemoryBuffer, ITensor
Inheritance
Object Platform::Object IInspectable TensorInt64Bit
Attributes
Implements

Windows requirements

Device family
Windows 10, version 1809 (introduced in 10.0.17763.0)
API contract
Windows.AI.MachineLearning.MachineLearningContract (introduced in v1.0)

Remarks

A tensor is a multi-dimensional array of values. A 64-bit signed integer tensor is a tensor of 64-bit signed integer values.

The layout of tensors is row-major, with tightly packed contiguous data representing each dimension. The total size of a tensor is the product of the size of each dimension.

Windows Server

To use this API on Windows Server, you must use Windows Server 2019 with Desktop Experience.

Thread safety

This API is thread-safe.

Version history

Windows version SDK version Value added
1903 18362 CreateFromBuffer
1903 18362 CreateFromShapeArrayAndDataArray

Properties

Kind

The kind of the feature.

Shape

Returns the count and size of each dimension.

TensorKind

Returns the data type of the tensor.

Methods

Close()

If there are no outstanding IMemoryBufferReference objects created from this object, Close will dispose of the object and destroy its associated resources. Otherwise, Close will detach itself from its associated resources and they will be destroyed when the last IMemoryBufferReference is closed.

Create()

Creates a 64-bit signed integer tensor object without allocating a buffer.

Create(IIterable<Int64>)

Creates a 64-bit signed integer tensor object and allocates a buffer of size shape.

CreateFromArray(IIterable<Int64>, Int64[])

Creates a 64-bit signed integer tensor object, allocates a buffer of size shape, and copies all of data into it.

CreateFromBuffer(Int64[], IBuffer)

Creates a 64-bit signed integer tensor object with the given shape and uses the underlying buffer in buffer for subsequent evaluation.

CreateFromIterable(IIterable<Int64>, IIterable<Int64>)

Creates a 64-bit signed integer tensor object, allocates a buffer of size shape, and copies all of data into it.

CreateFromShapeArrayAndDataArray(Int64[], Int64[])

Creates a 64-bit signed integer tensor object, allocates a buffer of size shape, and copies all of data into it.

CreateReference()

Returns an IMemoryBufferReference representation of the buffer backing the 64-bit signed integer tensor object.

Dispose()

Performs application-defined tasks associated with freeing, releasing, or resetting unmanaged resources.

GetAsVectorView()

Returns a read-only view of the data.

Applies to

See also