Custom operators

The Windows Machine Learning custom operator Win32 APIs are located in MLOperatorAuthor.h.

APIs

The following is a list of the custom operator APIs with their syntax and descriptions.

Enumerations

Name	Description
MLOperatorAttributeType	Specifies the type of an attribute. Each attribute type numerically matches the corresponding ONNX type.
MLOperatorEdgeType	Specifies the types of an input or output edge of an operator.
MLOperatorExecutionType	Specifies whether a kernel uses the CPU or GPU for computation.
MLOperatorKernelOptions	Specifies options used when registering custom operator kernels.
MLOperatorParameterOptions	Specifies option flags of input and output edges of operators.
MLOperatorSchemaEdgeTypeFormat	Specifies the manner in which types of input and output edges are described.
MLOperatorTensorDataType	Specifies the data type of a tensor. Each data type numerically matches the corresponding ONNX type.

Functions

Name	Description
MLCreateOperatorRegistry	Creates an instance of IMLOperatorRegistry which may be used to register a custom operator kernel and custom operator schema.

Interfaces

Name	Description
IMLOperatorAttributes	Represents the values of an operator's attributes, as determined by a model using the operator.
IMLOperatorKernel	Implemented by custom operator kernels.
IMLOperatorKernelContext	Provides information about an operator's usage while kernels are being computed.
IMLOperatorKernelCreationContext	Provides information about an operator's usage while kernels are being created.
IMLOperatorKernelFactory	Implemented by the author of a custom operator kernel to create instances of that kernel.
IMLOperatorRegistry	Represents an instance of a registry for the custom operator kernel and schema.
IMLOperatorShapeInferenceContext	Provides information about an operator's usage while shape inferrers are being invoked.
IMLOperatorShapeInferrer	Implemented by shape inferrers to infer shapes of an operator's output tensor edges.
IMLOperatorTensor	Representation of a tensor used during computation of custom operator kernels.
IMLOperatorTensorShapeDescription	Represents the set of input and output tensor shapes of an operator.
IMLOperatorTypeInferenceContext	Provides information about an operator's usage while type inferrers are being invoked.
IMLOperatorTypeInferrer	Implemented by type inferrers to infer the types of an operator's output edges.

Structures

Name	Description
MLOperatorAttribute	Specifies the name and properties of an attribute of a custom operator.
MLOperatorAttributeNameValue	Specifies the name and value(s) of an attribute of a custom operator.
MLOperatorEdgeDescription	Specifies the properties of an input or output edge of an operator.
MLOperatorEdgeTypeConstraint	Specifies constraints upon the types of edges supported in custom operator kernels and schema.
MLOperatorKernelDescription	Description of a custom operator kernel used to register that schema.
MLOperatorSchemaDescription	Description of a custom operator schema used to register that schema.
MLOperatorSchemaEdgeDescription	Specifies information about an input or output edge of an operator.
MLOperatorSetId	Specifies the identity of an operator set.

Note

Use the following resources for help with Windows ML:

• To ask or answer technical questions about Windows ML, please use the windows-machine-learning tag on Stack Overflow.
• To report a bug, please file an issue on our GitHub.