CoreML

Models 5

Load a trained machine learning model and inspect or extend its behavior.

• Cl
  MLModeliOS 11.0+

  MLModel
• Cl
  MLModelAssetiOS 16.0+

  A compiled model asset.
• Cl
  MLModelCollection

  MLModelCollection
• Pr
  MLCustomModeliOS 12.0+

  A protocol you implement to provide a model with your own prediction logic.
• Pr
  MLWritableiOS 13.0+

  A protocol for model types that can write themselves to a location on disk.

Model Description and Metadata 5

Examine a model's inputs, outputs, parameters, and descriptive metadata.

• Cl
  MLModelDescriptioniOS 11.0+

  A description of a model containing input, output, and state feature descriptions, optionally outputted features
• Cl
  MLFeatureDescriptioniOS 11.0+

  Description of a feature
• Cl
  MLParameterDescriptioniOS 13.0+

  Describes a model parameter along with a default value and any applicable constaint on the values.
• St
  MLModelMetadataKey

  Keys to a dictionary that holds useful information about a model.
• En
  MLModelStructureiOS 17.4+

  An enum representing the structure of a model.

Feature Values and Providers 5

Represent and supply the feature values a model consumes and produces.

• Cl
  MLFeatureValueiOS 11.0+

  An immutable variant holding a data value of a supported MLFeatureType
• St
  MLSendableFeatureValueiOS 18.0+

  A sendable feature value.
• Pr
  MLFeatureProvideriOS 11.0+

  A protocol that supplies named feature values as a model's inputs or outputs.
• Cl
  MLDictionaryFeatureProvideriOS 11.0+

  A concrete convenience class conforming to MLFeatureProvider.
• En
  MLFeatureTypeiOS 11.0+

  Supported data type enumeration

Multidimensional Arrays and Tensors 13

Hold numeric, sequence, and tensor data passed through a model.

• Cl
  MLMultiArrayiOS 11.0+

  Use `MLMultiArray` to store a multi-dimensional value.
• En
  MLMultiArrayDataTypeiOS 11.0+

  The data type of scalars in the multi-array.
• Cl
  MLSequenceiOS 12.0+

  An immutable container holding an ordered collection of feature values
• Cl
  MLStateiOS 18.0+

  Handle to the state buffers.
• St
  MLTensoriOS 18.0+

  A multi-dimensional array of numerical or Boolean scalars tailored to ML use cases, containing methods to perform
• Pr
  MLTensorRangeExpressioniOS 18.0+

  A type that can be used to slice a dimension of a tensor. Don't use this type directly.
• Pr
  MLTensorScalariOS 18.0+

  A type that represents the tensor scalar types supported by the framework. Don't use this type directly.
• St
  MLShapedArrayiOS 15.0+

  MLShapedArray is a N-dimensional array with non-contiguous strides.
• St
  MLShapedArraySliceiOS 15.0+

  A view into a contiguous region of a shaped array.
• Pr
  MLShapedArrayProtocoliOS 15.0+

  Protocol for the shaped array and the slice of shaped array.
• Pr
  MLShapedArrayRangeExpressioniOS 15.0+

  Declares a set of operations a range expression (e.g. 0 ..< 10) must implement in order to
• Pr
  MLShapedArrayScalariOS 15.0+

  Declares a set of operations all the data types of MLShapedArray's element must implement.
• En
  MLShapedArrayBufferLayoutiOS 18.0+

  Buffer layout enum

Feature Constraints 11

Describe the size, shape, and value limits a model places on its features.

• Cl
  MLImageConstraintiOS 11.0+

  MLImageConstraint
• Cl
  MLImageSizeiOS 12.0+

  An object that represents a discrete pixel width and height a model supports.
• Cl
  MLImageSizeConstraintiOS 12.0+

  An object that describes the range or set of image sizes a model accepts.
• En
  MLImageSizeConstraintTypeiOS 12.0+

  Constants that specify how a model constrains acceptable image sizes.
• Cl
  MLMultiArrayConstraintiOS 11.0+

  Constraint describing expected MLMultiArray properties
• Cl
  MLMultiArrayShapeConstraintiOS 12.0+

  An object that describes the allowed shapes for a multiarray feature.
• En
  MLMultiArrayShapeConstraintTypeiOS 12.0+

  Constants that specify how a model constrains acceptable multiarray shapes.
• Cl
  MLDictionaryConstraintiOS 11.0+

  MLDictionaryConstraint
• Cl
  MLSequenceConstraintiOS 12.0+

  Constraint describing expected MLSequence properties
• Cl
  MLStateConstraintiOS 18.0+

  Constraint of a state feature value.
• Cl
  MLNumericConstraintiOS 13.0+

  Allows enforcement of constraints on the values of update parameters.

Batch Prediction 3

Run predictions over many inputs at once for higher throughput.

• Pr
  MLBatchProvideriOS 12.0+

  Protocol for accessing a collection of feature providers
• Cl
  MLArrayBatchProvideriOS 12.0+

  A concrete convenience class conforming to MLBatchProvider.
• Cl
  MLPredictionOptionsiOS 11.0+

  MLPredictionOptions

Configuration and Compute Devices 10

Control where and how a model runs across the available compute hardware.

• Cl
  MLModelConfigurationiOS 12.0+

  An object to hold options for loading a model.
• En
  MLComputeUnitsiOS 12.0+

  Constants that specify which compute units, such as CPU, GPU, or Neural Engine, a model may use.
• En
  MLComputeDeviceiOS 17.0+

  Represents a compute device capable of running machine learning computations
• Pr
  MLComputeDeviceProtocoliOS 17.0+

  Protocol for representing a compute device. The represented device
• Cl
  MLCPUComputeDeviceiOS 17.0+

  Represents a CPU compute device.
• Cl
  MLGPUComputeDeviceiOS 17.0+

  Represents a GPU compute device.
• Cl
  MLNeuralEngineComputeDeviceiOS 17.0+

  Represents a NeuralEngine compute device for inference of machine learning models.
• Cl
  MLComputePlaniOS 17.4+

  A class representing the compute plan of a model.
• St
  MLComputePolicyiOS 18.0+

  The compute policy determining what compute device, or compute devices, to execute ML workloads on.
• St
  MLOptimizationHintsiOS 17.4+

  A value that provides hints to optimize how a model is loaded and executed.

On-Device Model Updates 9

Personalize and retrain models on device, tracking progress and results.

• Cl
  MLUpdateTaskiOS 13.0+

  Main class for setting up and controlling a model update. It provides some utility class methods for performing an update synchronously as well as class constructors for configuring an update and give
• Cl
  MLUpdateContextiOS 13.0+

  Provides context for the update process when the progress or completion handlers are invoked.
• Cl
  MLUpdateProgressHandlersiOS 13.0+

  Allows applications to register for progress and completion handlers.
• St
  MLUpdateProgressEventiOS 13.0+

  Events on which update task is capable of invoking progress handler.
• Cl
  MLTaskiOS 13.0+

  Class that abstracts state transitions and basic task controls.
• En
  MLTaskStateiOS 13.0+

  All possible states an MLTask can be in.
• Cl
  MLKeyiOS 13.0+

  A class representing key used to store any value against
• Cl
  MLMetricKeyiOS 13.0+

  A class to specify list of supported model update metrics.
• Cl
  MLParameterKeyiOS 13.0+

  A class to specify list of supported model update parameters.

Custom Layers 1

Extend a model with your own layer implementations.

• Pr
  MLCustomLayeriOS 11.2+

  A protocol you implement to provide a neural network layer with your own computation.

Errors 1

Handle failures that occur while loading or running a model.

• St
  MLModelErroriOS 11.0+

  An error type that describes failures that occur while loading or running a model.

Extends 18