mlcompute

package
v0.19.0 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Aug 11, 2026 License: MIT Imports: 12 Imported by: 0

Documentation

Rendered for darwin/amd64

Overview

Package mlcompute provides a fluent Go API over the macOS MLCompute framework.

Construction

New… functions create objects; each With… method sets one property and returns its receiver, so configuration calls chain. A <Type>FromID constructor adopts an Objective-C object obtained elsewhere.

Lifecycle

A wrapper releases its Objective-C object automatically once the garbage collector finds it unreachable. Call Release to relinquish the reference deterministically (for objects holding scarce resources); Release is idempotent, and afterwards the wrapper's methods are no-ops returning zero values.

Types

Each base type below lists the concrete types you construct and pass where the base is accepted:

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func GetRNGseed added in v0.17.0 

func GetRNGseed() *foundation.Number

GetRNGseed returns the global random number generator seed value.

func MLCActivationTypeDebugDescription 

func MLCActivationTypeDebugDescription(activationType ActivationType) string

MLCActivationTypeDebugDescription calls the MLCompute framework function MLCActivationTypeDebugDescription.

func MLCArithmeticOperationDebugDescription 

func MLCArithmeticOperationDebugDescription(operation ArithmeticOperation) string

MLCArithmeticOperationDebugDescription calls the MLCompute framework function MLCArithmeticOperationDebugDescription.

func MLCComparisonOperationDebugDescription 

func MLCComparisonOperationDebugDescription(operation ComparisonOperation) string

MLCComparisonOperationDebugDescription calls the MLCompute framework function MLCComparisonOperationDebugDescription.

func MLCConvolutionTypeDebugDescription 

func MLCConvolutionTypeDebugDescription(convolutionType ConvolutionType) string

MLCConvolutionTypeDebugDescription calls the MLCompute framework function MLCConvolutionTypeDebugDescription.

func MLCGradientClippingTypeDebugDescription 

func MLCGradientClippingTypeDebugDescription(gradientClippingType GradientClippingType) string

MLCGradientClippingTypeDebugDescription calls the MLCompute framework function MLCGradientClippingTypeDebugDescription.

func MLCLSTMResultModeDebugDescription 

func MLCLSTMResultModeDebugDescription(mode LSTMResultMode) string

MLCLSTMResultModeDebugDescription calls the MLCompute framework function MLCLSTMResultModeDebugDescription.

func MLCLossTypeDebugDescription 

func MLCLossTypeDebugDescription(lossType LossType) string

MLCLossTypeDebugDescription calls the MLCompute framework function MLCLossTypeDebugDescription.

func MLCPaddingPolicyDebugDescription 

func MLCPaddingPolicyDebugDescription(paddingPolicy PaddingPolicy) string

MLCPaddingPolicyDebugDescription calls the MLCompute framework function MLCPaddingPolicyDebugDescription.

func MLCPaddingTypeDebugDescription 

func MLCPaddingTypeDebugDescription(paddingType PaddingType) string

MLCPaddingTypeDebugDescription calls the MLCompute framework function MLCPaddingTypeDebugDescription.

func MLCPoolingTypeDebugDescription 

func MLCPoolingTypeDebugDescription(poolingType PoolingType) string

MLCPoolingTypeDebugDescription calls the MLCompute framework function MLCPoolingTypeDebugDescription.

func MLCReductionTypeDebugDescription 

func MLCReductionTypeDebugDescription(reductionType ReductionType) string

MLCReductionTypeDebugDescription calls the MLCompute framework function MLCReductionTypeDebugDescription.

func MLCSampleModeDebugDescription 

func MLCSampleModeDebugDescription(mode SampleMode) string

MLCSampleModeDebugDescription calls the MLCompute framework function MLCSampleModeDebugDescription.

func MLCSoftmaxOperationDebugDescription 

func MLCSoftmaxOperationDebugDescription(operation SoftmaxOperation) string

MLCSoftmaxOperationDebugDescription calls the MLCompute framework function MLCSoftmaxOperationDebugDescription.

func MaxTensorDimensions added in v0.17.0 

func MaxTensorDimensions() int

MaxTensorDimensions returns the maximum number of tensor dimensions supported

func SetRNGSeedTo added in v0.17.0 

func SetRNGSeedTo(seed obj.Object)

SetRNGSeedTo sets the global random number generator seed value.

func SupportsDataTypeOnDevice added in v0.17.0 

func SupportsDataTypeOnDevice(dataType DataType, device *Device) bool

SupportsDataTypeOnDevice returns a Boolean that indicates whether instances of this layer accept source tensors for the data type and device that you specify.

Types

type ActivationDescriptor added in v0.17.0 

type ActivationDescriptor struct {
	objref.Handle
}

ActivationDescriptor is an idiomatic wrapper over the Objective-C class MLCActivationDescriptor.

A configuration object you use to create an activation layer.

func ActivationDescriptorFromID added in v0.17.0 

func ActivationDescriptorFromID(id objc.ID) *ActivationDescriptor

ActivationDescriptorFromID adopts an existing Objective-C object as a ActivationDescriptor (nil for 0), retaining it and registering a release finalizer.

func DescriptorWithType added in v0.17.0 

func DescriptorWithType(activationType ActivationType) *ActivationDescriptor

DescriptorWithType creates an activation descriptor with the activation type you specify.

func DescriptorWithTypeA added in v0.17.0 

func DescriptorWithTypeA(activationType ActivationType, a float32) *ActivationDescriptor

DescriptorWithTypeA creates an activation descriptor with the activation type and parameter a that you specify.

func DescriptorWithTypeAB added in v0.17.0 

func DescriptorWithTypeAB(activationType ActivationType, a float32, b float32) *ActivationDescriptor

DescriptorWithTypeAB creates an activation descriptor with the activation type and parameters a and b that you specify.

func DescriptorWithTypeABC added in v0.17.0 

func DescriptorWithTypeABC(activationType ActivationType, a float32, b float32, c float32) *ActivationDescriptor

DescriptorWithTypeABC creates an activation descriptor with the activation type and parameters a, b, and c that you specify.

func NewActivationDescriptor added in v0.17.0 

func NewActivationDescriptor() *ActivationDescriptor

NewActivationDescriptor creates a new ActivationDescriptor.

func (*ActivationDescriptor) A added in v0.17.0 

func (ad *ActivationDescriptor) A() float32

A returns parameter to the activation function

func (*ActivationDescriptor) ActivationType added in v0.17.0 

func (ad *ActivationDescriptor) ActivationType() ActivationType

ActivationType returns the type of activation function

func (*ActivationDescriptor) B added in v0.17.0 

func (ad *ActivationDescriptor) B() float32

B returns parameter to the activation function

func (*ActivationDescriptor) C added in v0.17.0 

func (ad *ActivationDescriptor) C() float32

C returns parameter to the activation function

func (*ActivationDescriptor) Description added in v0.17.0 

func (ad *ActivationDescriptor) Description() string

Description returns the object's -description text.

func (*ActivationDescriptor) IsEqual added in v0.17.0 

func (ad *ActivationDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*ActivationDescriptor) IsKind added in v0.17.0 

func (ad *ActivationDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*ActivationDescriptor) String added in v0.17.0 

func (ad *ActivationDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

type ActivationLayer added in v0.17.0 

type ActivationLayer struct {
	Layer
}

ActivationLayer is an idiomatic wrapper over the Objective-C class MLCActivationLayer.

It embeds Layer, promoting that type's methods.

A layer that applies an activation function to the source tensor and produces an output.

func AbsoluteLayer added in v0.17.0 

func AbsoluteLayer() *ActivationLayer

AbsoluteLayer create an absolute activation layer

func ActivationLayerFromID added in v0.17.0 

func ActivationLayerFromID(id objc.ID) *ActivationLayer

ActivationLayerFromID adopts an existing Objective-C object as a ActivationLayer (nil for 0), retaining it and registering a release finalizer.

func CeluLayer added in v0.17.0 

func CeluLayer() *ActivationLayer

CeluLayer create a CELU activation layer

func CeluLayerWithA added in v0.17.0 

func CeluLayerWithA(a float32) *ActivationLayer

CeluLayerWithA create a CELU activation layer

func ClampLayerWithMinValueMaxValue added in v0.17.0 

func ClampLayerWithMinValueMaxValue(minValue float32, maxValue float32) *ActivationLayer

ClampLayerWithMinValueMaxValue create a clamp activation layer

func EluLayer added in v0.17.0 

func EluLayer() *ActivationLayer

EluLayer create an ELU activation layer

func EluLayerWithA added in v0.17.0 

func EluLayerWithA(a float32) *ActivationLayer

EluLayerWithA create an ELU activation layer

func GeluLayer added in v0.17.0 

func GeluLayer() *ActivationLayer

GeluLayer create a GELU activation layer

func HardShrinkLayer added in v0.17.0 

func HardShrinkLayer() *ActivationLayer

HardShrinkLayer create a hard shrink activation layer

func HardShrinkLayerWithA added in v0.17.0 

func HardShrinkLayerWithA(a float32) *ActivationLayer

HardShrinkLayerWithA create a hard shrink activation layer

func HardSigmoidLayer added in v0.17.0 

func HardSigmoidLayer() *ActivationLayer

HardSigmoidLayer create a hard sigmoid activation layer

func HardSwishLayer added in v0.17.0 

func HardSwishLayer() *ActivationLayer

HardSwishLayer create a hardswish activation layer

func LayerWithDescriptor added in v0.17.0 

func LayerWithDescriptor(descriptor *ActivationDescriptor) *ActivationLayer

LayerWithDescriptor creates an activation layer with the descriptor you specify.

func LeakyReLULayer added in v0.17.0 

func LeakyReLULayer() *ActivationLayer

LeakyReLULayer create a leaky ReLU activation layer

func LeakyReLULayerWithNegativeSlope added in v0.17.0 

func LeakyReLULayerWithNegativeSlope(negativeSlope float32) *ActivationLayer

LeakyReLULayerWithNegativeSlope create a leaky ReLU activation layer

func LinearLayerWithScaleBias added in v0.17.0 

func LinearLayerWithScaleBias(scale float32, bias float32) *ActivationLayer

LinearLayerWithScaleBias create a linear activation layer

func LogSigmoidLayer added in v0.17.0 

func LogSigmoidLayer() *ActivationLayer

LogSigmoidLayer create a log sigmoid activation layer

func NewActivationLayer added in v0.17.0 

func NewActivationLayer() *ActivationLayer

NewActivationLayer creates a new ActivationLayer.

func Relu6Layer added in v0.17.0 

func Relu6Layer() *ActivationLayer

Relu6Layer create a ReLU6 activation layer

func ReluLayer added in v0.17.0 

func ReluLayer() *ActivationLayer

ReluLayer create a ReLU activation layer

func RelunLayerWithAB added in v0.17.0 

func RelunLayerWithAB(a float32, b float32) *ActivationLayer

RelunLayerWithAB create a ReLUN activation layer This can be used to implement layers such as ReLU6 for example.

func SeluLayer added in v0.17.0 

func SeluLayer() *ActivationLayer

SeluLayer create a SELU activation layer

func SigmoidLayer added in v0.17.0 

func SigmoidLayer() *ActivationLayer

SigmoidLayer create a sigmoid activation layer

func SoftPlusLayer added in v0.17.0 

func SoftPlusLayer() *ActivationLayer

SoftPlusLayer create a soft plus activation layer

func SoftPlusLayerWithBeta added in v0.17.0 

func SoftPlusLayerWithBeta(beta float32) *ActivationLayer

SoftPlusLayerWithBeta create a soft plus activation layer

func SoftShrinkLayer added in v0.17.0 

func SoftShrinkLayer() *ActivationLayer

SoftShrinkLayer create a soft shrink activation layer

func SoftShrinkLayerWithA added in v0.17.0 

func SoftShrinkLayerWithA(a float32) *ActivationLayer

SoftShrinkLayerWithA create a soft shrink activation layer

func SoftSignLayer added in v0.17.0 

func SoftSignLayer() *ActivationLayer

SoftSignLayer create a soft sign activation layer

func TanhLayer added in v0.17.0 

func TanhLayer() *ActivationLayer

TanhLayer create a tanh activation layer

func TanhShrinkLayer added in v0.17.0 

func TanhShrinkLayer() *ActivationLayer

TanhShrinkLayer create a TanhShrink activation layer

func ThresholdLayerWithThresholdReplacement added in v0.17.0 

func ThresholdLayerWithThresholdReplacement(threshold float32, replacement float32) *ActivationLayer

ThresholdLayerWithThresholdReplacement create a threshold activation layer

func (*ActivationLayer) Descriptor added in v0.17.0 

func (al *ActivationLayer) Descriptor() *ActivationDescriptor

Descriptor returns the activation descriptor

func (*ActivationLayer) WithIsDebuggingEnabled added in v0.17.0 

func (al *ActivationLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *ActivationLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*ActivationLayer) WithLabel added in v0.17.0 

func (al *ActivationLayer) WithLabel(label string) *ActivationLayer

WithLabel sets a string that helps identify this layer.

type ActivationType added in v0.17.0 

type ActivationType int32

An activation type that you specify for an activation descriptor. 

const (
	// An activation type that implements the identity function.
	ActivationTypeNone ActivationType = 0
	// An activation type that implements the rectified linear unit activation function.
	ActivationTypeReLU ActivationType = 1
	// An activation type that implements the linear activation function.
	ActivationTypeLinear ActivationType = 2
	// An activation type that implements the sigmoid activation function.
	ActivationTypeSigmoid ActivationType = 3
	// An activation type that implements the hard sigmoid activation function.
	ActivationTypeHardSigmoid ActivationType = 4
	// An activation type that implements the hyperbolic tangent activation function.
	ActivationTypeTanh ActivationType = 5
	// An activation type that implements the absolute activation function.
	ActivationTypeAbsolute ActivationType = 6
	// An activation type that implements the soft plus activation function.
	ActivationTypeSoftPlus ActivationType = 7
	// An activation type that implements the parametric soft sign activation function.
	ActivationTypeSoftSign ActivationType = 8
	// An activation type that implements the exponential linear unit activation function.
	ActivationTypeELU ActivationType = 9
	// An activation type that implements the ReLUN activation function.
	ActivationTypeReLUN ActivationType = 10
	// An activation type that implements the log sigmoid activation function.
	ActivationTypeLogSigmoid ActivationType = 11
	// An activation type that implements the scaled exponential linear unit activation function.
	ActivationTypeSELU ActivationType = 12
	// An activation type that implements the CELU activation function.
	ActivationTypeCELU ActivationType = 13
	// An activation type that implements the hard shrink activation function.
	ActivationTypeHardShrink ActivationType = 14
	// An activation type that implements the soft shrink activation function.
	ActivationTypeSoftShrink ActivationType = 15
	// An activation type that implements the hyperbolic tangent shrink activation function.
	ActivationTypeTanhShrink ActivationType = 16
	// An activation type that implements the threshold activation function.
	ActivationTypeThreshold ActivationType = 17
	// An activation type that implements the gaussian error linear unit activation function.
	ActivationTypeGELU ActivationType = 18
	// An activation type that implements the hard swish activation function.
	ActivationTypeHardSwish ActivationType = 19
	// An activation type that implements the clamp activation function.
	ActivationTypeClamp ActivationType = 20
	// The count of activation types.
	ActivationTypeCount ActivationType = 21
)

func (ActivationType) String added in v0.17.0 

func (e ActivationType) String() string

String returns the ActivationType constant's name, or its numeric form when the value is not a known constant.

type AdamOptimizer added in v0.17.0 

type AdamOptimizer struct {
	Optimizer
}

AdamOptimizer is an idiomatic wrapper over the Objective-C class MLCAdamOptimizer.

It embeds Optimizer, promoting that type's methods.

An optimizer that represents the adaptive moment estimation algorithm.

func AdamOptimizerFromID added in v0.17.0 

func AdamOptimizerFromID(id objc.ID) *AdamOptimizer

AdamOptimizerFromID adopts an existing Objective-C object as a AdamOptimizer (nil for 0), retaining it and registering a release finalizer.

func NewAdamOptimizer added in v0.17.0 

func NewAdamOptimizer() *AdamOptimizer

NewAdamOptimizer creates a new AdamOptimizer.

func OptimizerWithDescriptor added in v0.17.0 

func OptimizerWithDescriptor(optimizerDescriptor *OptimizerDescriptor) *AdamOptimizer

OptimizerWithDescriptor creates an Adam optimizer with the descriptor you specify.

func OptimizerWithDescriptorBeta1Beta2EpsilonTimeStep added in v0.17.0 

func OptimizerWithDescriptorBeta1Beta2EpsilonTimeStep(optimizerDescriptor *OptimizerDescriptor, beta1 float32, beta2 float32, epsilon float32, timeStep int) *AdamOptimizer

OptimizerWithDescriptorBeta1Beta2EpsilonTimeStep creates an Adam optimizer with the values you specify.

func OptimizerWithDescriptorBeta1Beta2EpsilonUsesAMSGradTimeStep added in v0.17.0 

func OptimizerWithDescriptorBeta1Beta2EpsilonUsesAMSGradTimeStep(optimizerDescriptor *OptimizerDescriptor, beta1 float32, beta2 float32, epsilon float32, usesAMSGrad bool, timeStep int) *AdamOptimizer

OptimizerWithDescriptorBeta1Beta2EpsilonUsesAMSGradTimeStep creates an Adam optimizer with the values you specify.

func (*AdamOptimizer) Beta1 added in v0.17.0 

func (ao *AdamOptimizer) Beta1() float32

Beta1 returns coefficent used for computing running averages of gradient. The default is 0.9.

func (*AdamOptimizer) Beta2 added in v0.17.0 

func (ao *AdamOptimizer) Beta2() float32

Beta2 returns coefficent used for computing running averages of square of gradient. The default is 0.999.

func (*AdamOptimizer) Epsilon added in v0.17.0 

func (ao *AdamOptimizer) Epsilon() float32

Epsilon returns the epsilon.

func (*AdamOptimizer) TimeStep added in v0.17.0 

func (ao *AdamOptimizer) TimeStep() int

TimeStep returns the current timestep used for the update. The default is 1.

func (*AdamOptimizer) UsesAMSGrad added in v0.17.0 

func (ao *AdamOptimizer) UsesAMSGrad() bool

UsesAMSGrad reports whether to use the AMSGrad variant of this algorithm The default is false

func (*AdamOptimizer) WithAppliesGradientClipping added in v0.17.0 

func (ao *AdamOptimizer) WithAppliesGradientClipping(appliesGradientClipping bool) *AdamOptimizer

WithAppliesGradientClipping sets a Boolean value that indicates whether you apply gradient clipping.

func (*AdamOptimizer) WithLearningRate added in v0.17.0 

func (ao *AdamOptimizer) WithLearningRate(learningRate float32) *AdamOptimizer

WithLearningRate sets the learning rate.

type AdamWOptimizer added in v0.17.0 

type AdamWOptimizer struct {
	Optimizer
}

AdamWOptimizer is an idiomatic wrapper over the Objective-C class MLCAdamWOptimizer.

It embeds Optimizer, promoting that type's methods.

An optimizer that represents the Adam algorithm with weight decay.

func AdamWOptimizerFromID added in v0.17.0 

func AdamWOptimizerFromID(id objc.ID) *AdamWOptimizer

AdamWOptimizerFromID adopts an existing Objective-C object as a AdamWOptimizer (nil for 0), retaining it and registering a release finalizer.

func MLCAdamWOptimizerOptimizerWithDescriptor 

func MLCAdamWOptimizerOptimizerWithDescriptor(optimizerDescriptor *OptimizerDescriptor) *AdamWOptimizer

MLCAdamWOptimizerOptimizerWithDescriptor creates a default optimizer with the descriptor you specify.

func MLCAdamWOptimizerOptimizerWithDescriptorBeta1Beta2EpsilonUsesAMSGradTimeStep 

func MLCAdamWOptimizerOptimizerWithDescriptorBeta1Beta2EpsilonUsesAMSGradTimeStep(optimizerDescriptor *OptimizerDescriptor, beta1 float32, beta2 float32, epsilon float32, usesAMSGrad bool, timeStep int) *AdamWOptimizer

MLCAdamWOptimizerOptimizerWithDescriptorBeta1Beta2EpsilonUsesAMSGradTimeStep creates an AdamW optimizer with the values you specify.

func NewAdamWOptimizer added in v0.17.0 

func NewAdamWOptimizer() *AdamWOptimizer

NewAdamWOptimizer creates a new AdamWOptimizer.

func (*AdamWOptimizer) Beta1 added in v0.17.0 

func (awo *AdamWOptimizer) Beta1() float32

Beta1 returns coefficent used for computing running averages of gradient. The default is 0.9.

func (*AdamWOptimizer) Beta2 added in v0.17.0 

func (awo *AdamWOptimizer) Beta2() float32

Beta2 returns coefficent used for computing running averages of square of gradient. The default is 0.999.

func (*AdamWOptimizer) Epsilon added in v0.17.0 

func (awo *AdamWOptimizer) Epsilon() float32

Epsilon returns a term added to improve numerical stability. The default is 1e-8.

func (*AdamWOptimizer) TimeStep added in v0.17.0 

func (awo *AdamWOptimizer) TimeStep() int

TimeStep returns the current timestep used for the update. The default is 1.

func (*AdamWOptimizer) UsesAMSGrad added in v0.17.0 

func (awo *AdamWOptimizer) UsesAMSGrad() bool

UsesAMSGrad reports whether to use the AMSGrad variant of this algorithm The default is false

func (*AdamWOptimizer) WithAppliesGradientClipping added in v0.17.0 

func (awo *AdamWOptimizer) WithAppliesGradientClipping(appliesGradientClipping bool) *AdamWOptimizer

WithAppliesGradientClipping sets a Boolean value that indicates whether you apply gradient clipping.

func (*AdamWOptimizer) WithLearningRate added in v0.17.0 

func (awo *AdamWOptimizer) WithLearningRate(learningRate float32) *AdamWOptimizer

WithLearningRate sets the learning rate.

type ArithmeticLayer added in v0.17.0 

type ArithmeticLayer struct {
	Layer
}

ArithmeticLayer is an idiomatic wrapper over the Objective-C class MLCArithmeticLayer.

It embeds Layer, promoting that type's methods.

A layer that performs an arithmetic operation.

func ArithmeticLayerFromID added in v0.17.0 

func ArithmeticLayerFromID(id objc.ID) *ArithmeticLayer

ArithmeticLayerFromID adopts an existing Objective-C object as a ArithmeticLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithOperation added in v0.17.0 

func LayerWithOperation(operation ArithmeticOperation) *ArithmeticLayer

LayerWithOperation creates an arithmetic layer with the operation you specify.

func NewArithmeticLayer added in v0.17.0 

func NewArithmeticLayer() *ArithmeticLayer

NewArithmeticLayer creates a new ArithmeticLayer.

func (*ArithmeticLayer) Operation added in v0.17.0 

func (al *ArithmeticLayer) Operation() ArithmeticOperation

Operation returns the arithmetic operation.

func (*ArithmeticLayer) WithIsDebuggingEnabled added in v0.17.0 

func (al *ArithmeticLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *ArithmeticLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*ArithmeticLayer) WithLabel added in v0.17.0 

func (al *ArithmeticLayer) WithLabel(label string) *ArithmeticLayer

WithLabel sets a string that helps identify this layer.

type ArithmeticOperation added in v0.17.0 

type ArithmeticOperation int32

Constants that describe an arithmetic operation. 

const (
	// Calculates the element-wise sum of the inputs.
	ArithmeticOperationAdd ArithmeticOperation = 0
	// Calculates the element-wise difference between the inputs.
	ArithmeticOperationSubtract ArithmeticOperation = 1
	// Calculates the element-wise product of the inputs.
	ArithmeticOperationMultiply ArithmeticOperation = 2
	// Calculates the element-wise division of the inputs.
	ArithmeticOperationDivide ArithmeticOperation = 3
	// Calculates the element-wise floor of the inputs.
	ArithmeticOperationFloor ArithmeticOperation = 4
	// Calculates the element-wise rounding of the inputs.
	ArithmeticOperationRound ArithmeticOperation = 5
	// Calculates the element-wise ceiling of the inputs.
	ArithmeticOperationCeil ArithmeticOperation = 6
	// Calculates the element-wise square root of the input.
	ArithmeticOperationSqrt ArithmeticOperation = 7
	// Calculates the element-wise reciprocal of the square root of the input.
	ArithmeticOperationRsqrt ArithmeticOperation = 8
	// Calculates the element-wise sine of the input.
	ArithmeticOperationSin ArithmeticOperation = 9
	// Calculates the element-wise cosine of the input.
	ArithmeticOperationCos ArithmeticOperation = 10
	// Calculates the element-wise tangent of the input.
	ArithmeticOperationTan ArithmeticOperation = 11
	// Calculates the element-wise inverse sine of the input.
	ArithmeticOperationAsin ArithmeticOperation = 12
	// Calculates the element-wise inverse cosine of the input.
	ArithmeticOperationAcos ArithmeticOperation = 13
	// Calculates the element-wise inverse tangent of the input.
	ArithmeticOperationAtan ArithmeticOperation = 14
	// Calculates the element-wise hyperbolic sine of the input.
	ArithmeticOperationSinh ArithmeticOperation = 15
	// Calculates the element-wise hyperbolic cosine of the input.
	ArithmeticOperationCosh ArithmeticOperation = 16
	// Calculates the element-wise hyperbolic tangent of the input.
	ArithmeticOperationTanh ArithmeticOperation = 17
	// Calculates the element-wise inverse hyperbolic sine of the input.
	ArithmeticOperationAsinh ArithmeticOperation = 18
	// Calculates the element-wise inverse hyperbolic cosine of the input.
	ArithmeticOperationAcosh ArithmeticOperation = 19
	// Calculates the element-wise inverse hyperbolic tangent of the input.
	ArithmeticOperationAtanh ArithmeticOperation = 20
	// Calculates the element-wise first input raised to the power of the second input.
	ArithmeticOperationPow ArithmeticOperation = 21
	// Calculates the element-wise result of the exponent raised to the power of the input.
	ArithmeticOperationExp ArithmeticOperation = 22
	// Calculates the element-wise result of the number 2 raised to the power of the input.
	ArithmeticOperationExp2 ArithmeticOperation = 23
	// Calculates the element-wise natural logarithm of the input.
	ArithmeticOperationLog ArithmeticOperation = 24
	// Calculates the element-wise base 2 logarithm of the input.
	ArithmeticOperationLog2 ArithmeticOperation = 25
	// Calculates the element-wise product of the inputs, and returns 0 when the result isn’t a number or infinity.
	ArithmeticOperationMultiplyNoNaN ArithmeticOperation = 26
	// Calculates the element-wise division of the inputs, and returns 0 if the denominator is 0.
	ArithmeticOperationDivideNoNaN ArithmeticOperation = 27
	// Calculates the element-wise minimum of the inputs.
	ArithmeticOperationMin ArithmeticOperation = 28
	// Calculates the element-wise maximum the inputs.
	ArithmeticOperationMax ArithmeticOperation = 29
	// The total number of arithmetic operations.
	ArithmeticOperationCount ArithmeticOperation = 30
)

func (ArithmeticOperation) String added in v0.17.0 

func (e ArithmeticOperation) String() string

String returns the ArithmeticOperation constant's name, or its numeric form when the value is not a known constant.

type BatchNormalizationLayer added in v0.17.0 

type BatchNormalizationLayer struct {
	Layer
}

BatchNormalizationLayer is an idiomatic wrapper over the Objective-C class MLCBatchNormalizationLayer.

It embeds Layer, promoting that type's methods.

A layer that normalizes a batch of inputs.

func BatchNormalizationLayerFromID added in v0.17.0 

func BatchNormalizationLayerFromID(id objc.ID) *BatchNormalizationLayer

BatchNormalizationLayerFromID adopts an existing Objective-C object as a BatchNormalizationLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithFeatureChannelCountMeanVarianceBetaGammaVarianceEpsilon added in v0.17.0 

func LayerWithFeatureChannelCountMeanVarianceBetaGammaVarianceEpsilon(featureChannelCount int, mean *Tensor, variance *Tensor, beta *Tensor, gamma *Tensor, varianceEpsilon float32) *BatchNormalizationLayer

LayerWithFeatureChannelCountMeanVarianceBetaGammaVarianceEpsilon creates a batch normalization layer with the number of feature channels, tensors, and variance epsilon you specify.

func LayerWithFeatureChannelCountMeanVarianceBetaGammaVarianceEpsilonMomentum added in v0.17.0 

func LayerWithFeatureChannelCountMeanVarianceBetaGammaVarianceEpsilonMomentum(featureChannelCount int, mean *Tensor, variance *Tensor, beta *Tensor, gamma *Tensor, varianceEpsilon float32, momentum float32) *BatchNormalizationLayer

LayerWithFeatureChannelCountMeanVarianceBetaGammaVarianceEpsilonMomentum creates a batch normalization layer with the number of feature channels, tensors, variance epsilon, and momentum you specify.

func NewBatchNormalizationLayer added in v0.17.0 

func NewBatchNormalizationLayer() *BatchNormalizationLayer

NewBatchNormalizationLayer creates a new BatchNormalizationLayer.

func (*BatchNormalizationLayer) Beta added in v0.17.0 

func (bnl *BatchNormalizationLayer) Beta() *Tensor

Beta returns the beta tensor

func (*BatchNormalizationLayer) BetaParameter added in v0.17.0 

func (bnl *BatchNormalizationLayer) BetaParameter() *TensorParameter

BetaParameter returns the beta tensor parameter used for optimizer update

func (*BatchNormalizationLayer) FeatureChannelCount added in v0.17.0 

func (bnl *BatchNormalizationLayer) FeatureChannelCount() int

FeatureChannelCount returns the number of feature channels

func (*BatchNormalizationLayer) Gamma added in v0.17.0 

func (bnl *BatchNormalizationLayer) Gamma() *Tensor

Gamma returns the gamma tensor

func (*BatchNormalizationLayer) GammaParameter added in v0.17.0 

func (bnl *BatchNormalizationLayer) GammaParameter() *TensorParameter

GammaParameter returns the gamma tensor parameter used for optimizer update

func (*BatchNormalizationLayer) Mean added in v0.17.0 

func (bnl *BatchNormalizationLayer) Mean() *Tensor

Mean returns the mean tensor

func (*BatchNormalizationLayer) Momentum added in v0.17.0 

func (bnl *BatchNormalizationLayer) Momentum() float32

Momentum returns the value used for the running mean and variance computation The default is 0.99f.

func (*BatchNormalizationLayer) Variance added in v0.17.0 

func (bnl *BatchNormalizationLayer) Variance() *Tensor

Variance returns the variance tensor

func (*BatchNormalizationLayer) VarianceEpsilon added in v0.17.0 

func (bnl *BatchNormalizationLayer) VarianceEpsilon() float32

VarianceEpsilon returns a value used for numerical stability

func (*BatchNormalizationLayer) WithIsDebuggingEnabled added in v0.17.0 

func (bnl *BatchNormalizationLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *BatchNormalizationLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*BatchNormalizationLayer) WithLabel added in v0.17.0 

func (bnl *BatchNormalizationLayer) WithLabel(label string) *BatchNormalizationLayer

WithLabel sets a string that helps identify this layer.

type Clockid added in v0.17.0 

type Clockid int32
const (
	ClockidRealtime           Clockid = 0
	ClockidMonotonic          Clockid = 6
	ClockidMonotonicRaw       Clockid = 4
	ClockidMonotonicRawApprox Clockid = 5
	ClockidUptimeRaw          Clockid = 8
	ClockidUptimeRawApprox    Clockid = 9
	ClockidProcessCputimeID   Clockid = 12
	ClockidThreadCputimeID    Clockid = 16
)

func (Clockid) String added in v0.17.0 

func (e Clockid) String() string

String returns the Clockid constant's name, or its numeric form when the value is not a known constant.

type ComparisonLayer added in v0.17.0 

type ComparisonLayer struct {
	Layer
}

ComparisonLayer is an idiomatic wrapper over the Objective-C class MLCComparisonLayer.

It embeds Layer, promoting that type's methods.

A layer that performs elementwise comparison of two tensors.

func ComparisonLayerFromID added in v0.17.0 

func ComparisonLayerFromID(id objc.ID) *ComparisonLayer

ComparisonLayerFromID adopts an existing Objective-C object as a ComparisonLayer (nil for 0), retaining it and registering a release finalizer.

func MLCComparisonLayerLayerWithOperation 

func MLCComparisonLayerLayerWithOperation(operation ComparisonOperation) *ComparisonLayer

MLCComparisonLayerLayerWithOperation creates a comparison layer with the operation you specify.

func NewComparisonLayer added in v0.17.0 

func NewComparisonLayer() *ComparisonLayer

NewComparisonLayer creates a new ComparisonLayer.

func (*ComparisonLayer) Operation added in v0.17.0 

func (cl *ComparisonLayer) Operation() ComparisonOperation

Operation returns the operation.

func (*ComparisonLayer) WithIsDebuggingEnabled added in v0.17.0 

func (cl *ComparisonLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *ComparisonLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*ComparisonLayer) WithLabel added in v0.17.0 

func (cl *ComparisonLayer) WithLabel(label string) *ComparisonLayer

WithLabel sets a string that helps identify this layer.

type ComparisonOperation added in v0.17.0 

type ComparisonOperation int32

A comparison operation. 

const (
	ComparisonOperationEqual          ComparisonOperation = 0
	ComparisonOperationNotEqual       ComparisonOperation = 1
	ComparisonOperationLess           ComparisonOperation = 2
	ComparisonOperationGreater        ComparisonOperation = 3
	ComparisonOperationLessOrEqual    ComparisonOperation = 4
	ComparisonOperationGreaterOrEqual ComparisonOperation = 5
	ComparisonOperationLogicalAND     ComparisonOperation = 6
	ComparisonOperationLogicalOR      ComparisonOperation = 7
	ComparisonOperationLogicalNOT     ComparisonOperation = 8
	ComparisonOperationLogicalNAND    ComparisonOperation = 9
	ComparisonOperationLogicalNOR     ComparisonOperation = 10
	ComparisonOperationLogicalXOR     ComparisonOperation = 11
	// A number that represents the operation count.
	ComparisonOperationCount ComparisonOperation = 12
)

func (ComparisonOperation) String added in v0.17.0 

func (e ComparisonOperation) String() string

String returns the ComparisonOperation constant's name, or its numeric form when the value is not a known constant.

type ConcatenationLayer added in v0.17.0 

type ConcatenationLayer struct {
	Layer
}

ConcatenationLayer is an idiomatic wrapper over the Objective-C class MLCConcatenationLayer.

It embeds Layer, promoting that type's methods.

A layer that combines tensors into a single tensor.

func ConcatenationLayerFromID added in v0.17.0 

func ConcatenationLayerFromID(id objc.ID) *ConcatenationLayer

ConcatenationLayerFromID adopts an existing Objective-C object as a ConcatenationLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithDimension added in v0.17.0 

func LayerWithDimension(dimension int) *ConcatenationLayer

LayerWithDimension creates a concatenation layer with the dimension you specify.

func MLCConcatenationLayerLayer 

func MLCConcatenationLayerLayer() *ConcatenationLayer

MLCConcatenationLayerLayer creates a concatenation layer with a dimension value of 1, which typically represents feature channels.

func NewConcatenationLayer added in v0.17.0 

func NewConcatenationLayer() *ConcatenationLayer

NewConcatenationLayer creates a new ConcatenationLayer.

func (*ConcatenationLayer) Dimension added in v0.17.0 

func (cl *ConcatenationLayer) Dimension() int

Dimension returns the dimension (or axis) along which to concatenate tensors The default value is 1 (which typically represents features channels)

func (*ConcatenationLayer) WithIsDebuggingEnabled added in v0.17.0 

func (cl *ConcatenationLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *ConcatenationLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*ConcatenationLayer) WithLabel added in v0.17.0 

func (cl *ConcatenationLayer) WithLabel(label string) *ConcatenationLayer

WithLabel sets a string that helps identify this layer.

type ConvolutionDescriptor added in v0.17.0 

type ConvolutionDescriptor struct {
	objref.Handle
}

ConvolutionDescriptor is an idiomatic wrapper over the Objective-C class MLCConvolutionDescriptor.

A configuration object you use to create a convolution or fully connected layer.

func ConvolutionDescriptorFromID added in v0.17.0 

func ConvolutionDescriptorFromID(id objc.ID) *ConvolutionDescriptor

ConvolutionDescriptorFromID adopts an existing Objective-C object as a ConvolutionDescriptor (nil for 0), retaining it and registering a release finalizer.

func ConvolutionTransposeDescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountGroupCountStridesDilationRatesPaddingPolicyPaddingSizes added in v0.17.0 

func ConvolutionTransposeDescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountGroupCountStridesDilationRatesPaddingPolicyPaddingSizes(kernelSizes []*foundation.Number, inputFeatureChannelCount int, outputFeatureChannelCount int, groupCount int, strides []*foundation.Number, dilationRates []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *ConvolutionDescriptor

ConvolutionTransposeDescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountGroupCountStridesDilationRatesPaddingPolicyPaddingSizes creates a convolution transpose descriptor with the kernel and padding options, number of feature channels and groups, and dilation rates you specify.

func ConvolutionTransposeDescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountStridesPaddingPolicyPaddingSizes added in v0.17.0 

func ConvolutionTransposeDescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountStridesPaddingPolicyPaddingSizes(kernelSizes []*foundation.Number, inputFeatureChannelCount int, outputFeatureChannelCount int, strides []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *ConvolutionDescriptor

ConvolutionTransposeDescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountStridesPaddingPolicyPaddingSizes creates a convolution transpose descriptor with the kernel sizes, number of feature channels, strides, and padding options you specify.

func ConvolutionTransposeDescriptorWithKernelWidthKernelHeightInputFeatureChannelCountOutputFeatureChannelCount added in v0.17.0 

func ConvolutionTransposeDescriptorWithKernelWidthKernelHeightInputFeatureChannelCountOutputFeatureChannelCount(kernelWidth int, kernelHeight int, inputFeatureChannelCount int, outputFeatureChannelCount int) *ConvolutionDescriptor

ConvolutionTransposeDescriptorWithKernelWidthKernelHeightInputFeatureChannelCountOutputFeatureChannelCount creates a descriptor for convolution transpose with the kernel sizes and number of feature channels you specify.

func DepthwiseConvolutionDescriptorWithKernelSizesInputFeatureChannelCountChannelMultiplierStridesDilationRatesPaddingPolicyPaddingSizes added in v0.17.0 

func DepthwiseConvolutionDescriptorWithKernelSizesInputFeatureChannelCountChannelMultiplierStridesDilationRatesPaddingPolicyPaddingSizes(kernelSizes []*foundation.Number, inputFeatureChannelCount int, channelMultiplier int, strides []*foundation.Number, dilationRates []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *ConvolutionDescriptor

DepthwiseConvolutionDescriptorWithKernelSizesInputFeatureChannelCountChannelMultiplierStridesDilationRatesPaddingPolicyPaddingSizes creates a convolution descriptor with the kernel and padding options, number of input channels, channel multiplier, and dilation rates you specify.

func DepthwiseConvolutionDescriptorWithKernelSizesInputFeatureChannelCountChannelMultiplierStridesPaddingPolicyPaddingSizes added in v0.17.0 

func DepthwiseConvolutionDescriptorWithKernelSizesInputFeatureChannelCountChannelMultiplierStridesPaddingPolicyPaddingSizes(kernelSizes []*foundation.Number, inputFeatureChannelCount int, channelMultiplier int, strides []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *ConvolutionDescriptor

DepthwiseConvolutionDescriptorWithKernelSizesInputFeatureChannelCountChannelMultiplierStridesPaddingPolicyPaddingSizes creates a depthwise convolution descriptor with the kernel and padding options, number of input feature channels, and channel multiplier you specify.

func DepthwiseConvolutionDescriptorWithKernelWidthKernelHeightInputFeatureChannelCountChannelMultiplier added in v0.17.0 

func DepthwiseConvolutionDescriptorWithKernelWidthKernelHeightInputFeatureChannelCountChannelMultiplier(kernelWidth int, kernelHeight int, inputFeatureChannelCount int, channelMultiplier int) *ConvolutionDescriptor

DepthwiseConvolutionDescriptorWithKernelWidthKernelHeightInputFeatureChannelCountChannelMultiplier creates a descriptor for depthwise convolution with the kernel sizes, number of input feature channels, and channel multiplier you specify.

func DescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountGroupCountStridesDilationRatesPaddingPolicyPaddingSizes added in v0.17.0 

func DescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountGroupCountStridesDilationRatesPaddingPolicyPaddingSizes(kernelSizes []*foundation.Number, inputFeatureChannelCount int, outputFeatureChannelCount int, groupCount int, strides []*foundation.Number, dilationRates []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *ConvolutionDescriptor

DescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountGroupCountStridesDilationRatesPaddingPolicyPaddingSizes creates a convolution descriptor with the kernel and padding options, number of feature channels and groups, and dilation rates you specify.

func DescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountStridesPaddingPolicyPaddingSizes added in v0.17.0 

func DescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountStridesPaddingPolicyPaddingSizes(kernelSizes []*foundation.Number, inputFeatureChannelCount int, outputFeatureChannelCount int, strides []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *ConvolutionDescriptor

DescriptorWithKernelSizesInputFeatureChannelCountOutputFeatureChannelCountStridesPaddingPolicyPaddingSizes creates a convolution descriptor with the kernel sizes, number of feature channels, strides, padding policy, and padding sizes you specify.

func DescriptorWithKernelWidthKernelHeightInputFeatureChannelCountOutputFeatureChannelCount added in v0.17.0 

func DescriptorWithKernelWidthKernelHeightInputFeatureChannelCountOutputFeatureChannelCount(kernelWidth int, kernelHeight int, inputFeatureChannelCount int, outputFeatureChannelCount int) *ConvolutionDescriptor

DescriptorWithKernelWidthKernelHeightInputFeatureChannelCountOutputFeatureChannelCount creates a convolution descriptor with the kernel sizes and number of feature channels you specify.

func DescriptorWithTypeKernelSizesInputFeatureChannelCountOutputFeatureChannelCountGroupCountStridesDilationRatesPaddingPolicyPaddingSizes added in v0.17.0 

func DescriptorWithTypeKernelSizesInputFeatureChannelCountOutputFeatureChannelCountGroupCountStridesDilationRatesPaddingPolicyPaddingSizes(convolutionType ConvolutionType, kernelSizes []*foundation.Number, inputFeatureChannelCount int, outputFeatureChannelCount int, groupCount int, strides []*foundation.Number, dilationRates []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *ConvolutionDescriptor

DescriptorWithTypeKernelSizesInputFeatureChannelCountOutputFeatureChannelCountGroupCountStridesDilationRatesPaddingPolicyPaddingSizes creates a descriptor with the type, kernel sizes, number of feature channels and groups, strides, dilation rates, and padding policy you specify.

func NewConvolutionDescriptor added in v0.17.0 

func NewConvolutionDescriptor() *ConvolutionDescriptor

NewConvolutionDescriptor creates a new ConvolutionDescriptor.

func (*ConvolutionDescriptor) ConvolutionType added in v0.17.0 

func (cd *ConvolutionDescriptor) ConvolutionType() ConvolutionType

ConvolutionType returns the type of convolution.

func (*ConvolutionDescriptor) Description added in v0.17.0 

func (cd *ConvolutionDescriptor) Description() string

Description returns the object's -description text.

func (*ConvolutionDescriptor) DilationRateInX added in v0.17.0 

func (cd *ConvolutionDescriptor) DilationRateInX() int

DilationRateInX returns the dilation rate i.e. stride of elements in the kernel in x.

func (*ConvolutionDescriptor) DilationRateInY added in v0.17.0 

func (cd *ConvolutionDescriptor) DilationRateInY() int

DilationRateInY returns the dilation rate i.e. stride of elements in the kernel in y.

func (*ConvolutionDescriptor) GroupCount added in v0.17.0 

func (cd *ConvolutionDescriptor) GroupCount() int

GroupCount returns number of blocked connections from input channels to output channels

func (*ConvolutionDescriptor) InputFeatureChannelCount added in v0.17.0 

func (cd *ConvolutionDescriptor) InputFeatureChannelCount() int

InputFeatureChannelCount returns number of channels in the input tensor

func (*ConvolutionDescriptor) IsConvolutionTranspose added in v0.17.0 

func (cd *ConvolutionDescriptor) IsConvolutionTranspose() bool

IsConvolutionTranspose reports whether a flag to indicate if this is a convolution transpose

func (*ConvolutionDescriptor) IsEqual added in v0.17.0 

func (cd *ConvolutionDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*ConvolutionDescriptor) IsKind added in v0.17.0 

func (cd *ConvolutionDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*ConvolutionDescriptor) KernelHeight added in v0.17.0 

func (cd *ConvolutionDescriptor) KernelHeight() int

KernelHeight returns the convolution kernel size in y.

func (*ConvolutionDescriptor) KernelWidth added in v0.17.0 

func (cd *ConvolutionDescriptor) KernelWidth() int

KernelWidth returns the convolution kernel size in x.

func (*ConvolutionDescriptor) OutputFeatureChannelCount added in v0.17.0 

func (cd *ConvolutionDescriptor) OutputFeatureChannelCount() int

OutputFeatureChannelCount returns number of channels in the output tensor

func (*ConvolutionDescriptor) PaddingPolicy added in v0.17.0 

func (cd *ConvolutionDescriptor) PaddingPolicy() PaddingPolicy

PaddingPolicy returns the padding policy to use.

func (*ConvolutionDescriptor) PaddingSizeInX added in v0.17.0 

func (cd *ConvolutionDescriptor) PaddingSizeInX() int

PaddingSizeInX returns the pooling size in x (left and right) to use if paddingPolicy is MLCPaddingPolicyUsePaddingSize

func (*ConvolutionDescriptor) PaddingSizeInY added in v0.17.0 

func (cd *ConvolutionDescriptor) PaddingSizeInY() int

PaddingSizeInY returns the pooling size in y (top and bottom) to use if paddingPolicy is MLCPaddingPolicyUsePaddingSize

func (*ConvolutionDescriptor) StrideInX added in v0.17.0 

func (cd *ConvolutionDescriptor) StrideInX() int

StrideInX returns the stride of the kernel in x.

func (*ConvolutionDescriptor) StrideInY added in v0.17.0 

func (cd *ConvolutionDescriptor) StrideInY() int

StrideInY returns the stride of the kernel in y.

func (*ConvolutionDescriptor) String added in v0.17.0 

func (cd *ConvolutionDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*ConvolutionDescriptor) UsesDepthwiseConvolution added in v0.17.0 

func (cd *ConvolutionDescriptor) UsesDepthwiseConvolution() bool

UsesDepthwiseConvolution reports whether a flag to indicate depthwise convolution

type ConvolutionLayer added in v0.17.0 

type ConvolutionLayer struct {
	Layer
}

ConvolutionLayer is an idiomatic wrapper over the Objective-C class MLCConvolutionLayer.

It embeds Layer, promoting that type's methods.

A layer that applies a convolution over a signal.

func ConvolutionLayerFromID added in v0.17.0 

func ConvolutionLayerFromID(id objc.ID) *ConvolutionLayer

ConvolutionLayerFromID adopts an existing Objective-C object as a ConvolutionLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithWeightsBiasesDescriptor added in v0.17.0 

func LayerWithWeightsBiasesDescriptor(weights *Tensor, biases *Tensor, descriptor *ConvolutionDescriptor) *ConvolutionLayer

LayerWithWeightsBiasesDescriptor creates a convolution layer with the weights, biases, and descriptor you specify.

func NewConvolutionLayer added in v0.17.0 

func NewConvolutionLayer() *ConvolutionLayer

NewConvolutionLayer creates a new ConvolutionLayer.

func (*ConvolutionLayer) Biases added in v0.17.0 

func (cl *ConvolutionLayer) Biases() *Tensor

Biases returns the bias tensor used by the convolution layer

func (*ConvolutionLayer) BiasesParameter added in v0.17.0 

func (cl *ConvolutionLayer) BiasesParameter() *TensorParameter

BiasesParameter returns the bias tensor parameter used for optimizer update

func (*ConvolutionLayer) Descriptor added in v0.17.0 

func (cl *ConvolutionLayer) Descriptor() *ConvolutionDescriptor

Descriptor returns the convolution descriptor

func (*ConvolutionLayer) Weights added in v0.17.0 

func (cl *ConvolutionLayer) Weights() *Tensor

Weights returns the weights tensor used by the convolution layer

func (*ConvolutionLayer) WeightsParameter added in v0.17.0 

func (cl *ConvolutionLayer) WeightsParameter() *TensorParameter

WeightsParameter returns the weights tensor parameter used for optimizer update

func (*ConvolutionLayer) WithIsDebuggingEnabled added in v0.17.0 

func (cl *ConvolutionLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *ConvolutionLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*ConvolutionLayer) WithLabel added in v0.17.0 

func (cl *ConvolutionLayer) WithLabel(label string) *ConvolutionLayer

WithLabel sets a string that helps identify this layer.

type ConvolutionType added in v0.17.0 

type ConvolutionType int32

The convolution type specified for a convolution layer. 

const (
	// The standard convolution type.
	ConvolutionTypeStandard ConvolutionType = 0
	// The transposed convolution type.
	ConvolutionTypeTransposed ConvolutionType = 1
	// The depthwise convolution type.
	ConvolutionTypeDepthwise ConvolutionType = 2
)

func (ConvolutionType) String added in v0.17.0 

func (e ConvolutionType) String() string

String returns the ConvolutionType constant's name, or its numeric form when the value is not a known constant.

type DataType added in v0.17.0 

type DataType int32

A tensor data type. 

const (
	DataTypeInvalid DataType = 0
	// The 32-bit floating-point data type.
	DataTypeFloat32 DataType = 1
	// The 16-bit floating-point data type.
	DataTypeFloat16 DataType = 3
	// The Boolean data type.
	DataTypeBoolean DataType = 4
	// The 64-bit integer data type.
	DataTypeInt64 DataType = 5
	// The 32-bit integer data type.
	DataTypeInt32 DataType = 7
	// The 8-bit integer data type.
	DataTypeInt8 DataType = 8
	// The 8-bit unsigned integer data type.
	DataTypeUInt8 DataType = 9
	DataTypeCount DataType = 10
)

func (DataType) String added in v0.17.0 

func (e DataType) String() string

String returns the DataType constant's name, or its numeric form when the value is not a known constant.

type Device added in v0.17.0 

type Device struct {
	objref.Handle
}

Device is an idiomatic wrapper over the Objective-C class MLCDevice.

An object that represents the CPU or one or more GPUs the framework uses to execute a neural network.

func AneDevice added in v0.17.0 

func AneDevice() *Device

AneDevice creates a device that uses the Apple Neural Engine, if one exists.

func CpuDevice added in v0.17.0 

func CpuDevice() *Device

CpuDevice creates a device that uses the CPU.

func DeviceFromID added in v0.17.0 

func DeviceFromID(id objc.ID) *Device

DeviceFromID adopts an existing Objective-C object as a Device (nil for 0), retaining it and registering a release finalizer.

func DeviceWithGPUDevices added in v0.17.0 

func DeviceWithGPUDevices(gpus []obj.Object) *Device

DeviceWithGPUDevices creates a device using the GPUs you specify.

func DeviceWithType added in v0.17.0 

func DeviceWithType(type_ DeviceType) *Device

DeviceWithType creates a device of the type you specify.

func DeviceWithTypeSelectsMultipleComputeDevices added in v0.17.0 

func DeviceWithTypeSelectsMultipleComputeDevices(type_ DeviceType, selectsMultipleComputeDevices bool) *Device

DeviceWithTypeSelectsMultipleComputeDevices creates a device that you can configure to use multiple compute devices.

func GpuDevice added in v0.17.0 

func GpuDevice() *Device

GpuDevice creates a device that uses a GPU, if one exists.

func NewDevice added in v0.17.0 

func NewDevice() *Device

NewDevice creates a new Device.

func (*Device) ActualDeviceType added in v0.17.0 

func (d *Device) ActualDeviceType() DeviceType

ActualDeviceType returns the specific device selected. This can be CPU, GPU or ANE. If type is MLCDeviceTypeAny, this property can be used to find out the specific device type that is selected.

func (*Device) Description added in v0.17.0 

func (d *Device) Description() string

Description returns the object's -description text.

func (*Device) GPUDevices added in v0.17.0 

func (d *Device) GPUDevices() []obj.Object

GPUDevices returns the GPU devices.

func (*Device) IsEqual added in v0.17.0 

func (d *Device) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*Device) IsKind added in v0.17.0 

func (d *Device) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*Device) String added in v0.17.0 

func (d *Device) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*Device) Type added in v0.17.0 

func (d *Device) Type() DeviceType

Type returns the type specified when the device is created Recommend that developers use MLCDeviceTypeAny as the device type. This will ensure that MLCompute will select the best device to execute the neural network. If developers want to be able to control device selection, they can select CPU or GPU and for the GPU, they can also select a specific Metal device.

type DeviceType added in v0.17.0 

type DeviceType int32

A device type for execution of a neural network. 

const (
	// A device type that represents the CPU.
	DeviceTypeCPU DeviceType = 0
	// A device type that represents the GPU.
	DeviceTypeGPU DeviceType = 1
	// A device type that represents either the CPU or GPU.
	DeviceTypeAny DeviceType = 2
	// A device type that represents the Apple Neural Engine.
	DeviceTypeANE DeviceType = 3
	// A number that represents the number of device types.
	DeviceTypeCount DeviceType = 4
)

func (DeviceType) String added in v0.17.0 

func (e DeviceType) String() string

String returns the DeviceType constant's name, or its numeric form when the value is not a known constant.

type DispatchAutoreleaseFrequency added in v0.17.0 

type DispatchAutoreleaseFrequency uint64
const (
	DispatchAutoreleaseFrequencyInherit  DispatchAutoreleaseFrequency = 0
	DispatchAutoreleaseFrequencyWorkItem DispatchAutoreleaseFrequency = 1
	DispatchAutoreleaseFrequencyNever    DispatchAutoreleaseFrequency = 2
)

func (DispatchAutoreleaseFrequency) String added in v0.17.0 

func (e DispatchAutoreleaseFrequency) String() string

String returns the DispatchAutoreleaseFrequency constant's name, or its numeric form when the value is not a known constant.

type DispatchBlockFlags added in v0.17.0 

type DispatchBlockFlags uint64

Bitmask — values may be combined with |. 

const (
	DispatchBlockFlagsBarrier         DispatchBlockFlags = 1
	DispatchBlockFlagsDetached        DispatchBlockFlags = 2
	DispatchBlockFlagsAssignCurrent   DispatchBlockFlags = 4
	DispatchBlockFlagsNoQosClass      DispatchBlockFlags = 8
	DispatchBlockFlagsInheritQosClass DispatchBlockFlags = 16
	DispatchBlockFlagsEnforceQosClass DispatchBlockFlags = 32
)

func (DispatchBlockFlags) String added in v0.17.0 

func (e DispatchBlockFlags) String() string

String returns the DispatchBlockFlags constant's name, or its numeric form when the value is not a known constant.

type DropoutLayer added in v0.17.0 

type DropoutLayer struct {
	Layer
}

DropoutLayer is an idiomatic wrapper over the Objective-C class MLCDropoutLayer.

It embeds Layer, promoting that type's methods.

A layer that deactivates neurons randomly to avoid overfitting.

func DropoutLayerFromID added in v0.17.0 

func DropoutLayerFromID(id objc.ID) *DropoutLayer

DropoutLayerFromID adopts an existing Objective-C object as a DropoutLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithRateSeed added in v0.17.0 

func LayerWithRateSeed(rate float32, seed int) *DropoutLayer

LayerWithRateSeed creates a dropout layer with the probability rate and random number generator seed you specify.

func NewDropoutLayer added in v0.17.0 

func NewDropoutLayer() *DropoutLayer

NewDropoutLayer creates a new DropoutLayer.

func (*DropoutLayer) Rate added in v0.17.0 

func (dl *DropoutLayer) Rate() float32

Rate returns the probability that each element is dropped

func (*DropoutLayer) Seed added in v0.17.0 

func (dl *DropoutLayer) Seed() int

Seed returns the initial seed used to generate random numbers

func (*DropoutLayer) WithIsDebuggingEnabled added in v0.17.0 

func (dl *DropoutLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *DropoutLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*DropoutLayer) WithLabel added in v0.17.0 

func (dl *DropoutLayer) WithLabel(label string) *DropoutLayer

WithLabel sets a string that helps identify this layer.

type EmbeddingDescriptor added in v0.17.0 

type EmbeddingDescriptor struct {
	objref.Handle
}

EmbeddingDescriptor is an idiomatic wrapper over the Objective-C class MLCEmbeddingDescriptor.

A configuration object you use to create an embedding layer.

func DescriptorWithEmbeddingCountEmbeddingDimension added in v0.17.0 

func DescriptorWithEmbeddingCountEmbeddingDimension(embeddingCount obj.Object, embeddingDimension obj.Object) *EmbeddingDescriptor

DescriptorWithEmbeddingCountEmbeddingDimension creates an embedding descriptor with the size of the dictionary and dimension of embedding vectors you specify.

func DescriptorWithEmbeddingCountEmbeddingDimensionPaddingIndexMaximumNormPNormScalesGradientByFrequency added in v0.17.0 

func DescriptorWithEmbeddingCountEmbeddingDimensionPaddingIndexMaximumNormPNormScalesGradientByFrequency(embeddingCount obj.Object, embeddingDimension obj.Object, paddingIndex obj.Object, maximumNorm obj.Object, pNorm obj.Object, scalesGradientByFrequency bool) *EmbeddingDescriptor

DescriptorWithEmbeddingCountEmbeddingDimensionPaddingIndexMaximumNormPNormScalesGradientByFrequency creates a new embedding descriptor with the size and dimension of embedding vectors, padding index, and norm and scaling options that you specify.

func EmbeddingDescriptorFromID added in v0.17.0 

func EmbeddingDescriptorFromID(id objc.ID) *EmbeddingDescriptor

EmbeddingDescriptorFromID adopts an existing Objective-C object as a EmbeddingDescriptor (nil for 0), retaining it and registering a release finalizer.

func NewEmbeddingDescriptor added in v0.17.0 

func NewEmbeddingDescriptor() *EmbeddingDescriptor

NewEmbeddingDescriptor creates a new EmbeddingDescriptor.

func (*EmbeddingDescriptor) Description added in v0.17.0 

func (ed *EmbeddingDescriptor) Description() string

Description returns the object's -description text.

func (*EmbeddingDescriptor) EmbeddingCount added in v0.17.0 

func (ed *EmbeddingDescriptor) EmbeddingCount() *foundation.Number

EmbeddingCount returns the size of the dictionary

func (*EmbeddingDescriptor) EmbeddingDimension added in v0.17.0 

func (ed *EmbeddingDescriptor) EmbeddingDimension() *foundation.Number

EmbeddingDimension returns the dimension of embedding vectors

func (*EmbeddingDescriptor) IsEqual added in v0.17.0 

func (ed *EmbeddingDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*EmbeddingDescriptor) IsKind added in v0.17.0 

func (ed *EmbeddingDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*EmbeddingDescriptor) MaximumNorm added in v0.17.0 

func (ed *EmbeddingDescriptor) MaximumNorm() *foundation.Number

MaximumNorm returns a float, if set, in the forward pass only, the selected embedding vectors will be re-normalized to have an Lp norm of less than maximumNorm in the dictionary, Default=nil

func (*EmbeddingDescriptor) PNorm added in v0.17.0 

func (ed *EmbeddingDescriptor) PNorm() *foundation.Number

PNorm returns a float, the p of the Lp norm, can be set to infinity norm by [NSNumber numberWithFloat:INFINITY]. Default=2.0

func (*EmbeddingDescriptor) PaddingIndex added in v0.17.0 

func (ed *EmbeddingDescriptor) PaddingIndex() *foundation.Number

PaddingIndex returns if set, the embedding vector at paddingIndex is initialized with zero and will not be updated in gradient pass, Default=nil

func (*EmbeddingDescriptor) ScalesGradientByFrequency added in v0.17.0 

func (ed *EmbeddingDescriptor) ScalesGradientByFrequency() bool

ScalesGradientByFrequency reports whether if set, the gradients are scaled by the inverse of the frequency of the words in batch before the weight update. Default=NO

func (*EmbeddingDescriptor) String added in v0.17.0 

func (ed *EmbeddingDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

type EmbeddingLayer added in v0.17.0 

type EmbeddingLayer struct {
	Layer
}

EmbeddingLayer is an idiomatic wrapper over the Objective-C class MLCEmbeddingLayer.

It embeds Layer, promoting that type's methods.

A layer that stores a word embedding.

func EmbeddingLayerFromID added in v0.17.0 

func EmbeddingLayerFromID(id objc.ID) *EmbeddingLayer

EmbeddingLayerFromID adopts an existing Objective-C object as a EmbeddingLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithDescriptorWeights added in v0.17.0 

func LayerWithDescriptorWeights(descriptor *EmbeddingDescriptor, weights *Tensor) *EmbeddingLayer

LayerWithDescriptorWeights creates an embedding layer with the descriptor and word embedding weights tensor you specify.

func NewEmbeddingLayer added in v0.17.0 

func NewEmbeddingLayer() *EmbeddingLayer

NewEmbeddingLayer creates a new EmbeddingLayer.

func (*EmbeddingLayer) Descriptor added in v0.17.0 

func (el *EmbeddingLayer) Descriptor() *EmbeddingDescriptor

Descriptor returns the descriptor.

func (*EmbeddingLayer) Weights added in v0.17.0 

func (el *EmbeddingLayer) Weights() *Tensor

Weights returns the array of word embeddings

func (*EmbeddingLayer) WeightsParameter added in v0.17.0 

func (el *EmbeddingLayer) WeightsParameter() *TensorParameter

WeightsParameter returns the weights tensor parameter used for optimizer update

func (*EmbeddingLayer) WithIsDebuggingEnabled added in v0.17.0 

func (el *EmbeddingLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *EmbeddingLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*EmbeddingLayer) WithLabel added in v0.17.0 

func (el *EmbeddingLayer) WithLabel(label string) *EmbeddingLayer

WithLabel sets a string that helps identify this layer.

type EntryID added in v0.17.0 

type EntryID int32
const (
	EntryIDFirstEntry EntryID = 0
	EntryIDNextEntry  EntryID = -1
	EntryIDLastEntry  EntryID = -2
)

func (EntryID) String added in v0.17.0 

func (e EntryID) String() string

String returns the EntryID constant's name, or its numeric form when the value is not a known constant.

type ExecutionOptions added in v0.17.0 

type ExecutionOptions uint64

A bitmask that specifies the options you use when executing a graph. Bitmask — values may be combined with |. 

const (
	// The option to execute the graph in the most efficient way possible.
	ExecutionOptionsNone ExecutionOptions = 0
	// The option to skip writing input data to device memory.
	ExecutionOptionsSkipWritingInputDataToDevice ExecutionOptions = 1
	// The option to execute the graph synchronously.
	ExecutionOptionsSynchronous ExecutionOptions = 2
	// The option to return profiling information in the callback before returning from execution.
	ExecutionOptionsProfiling ExecutionOptions = 4
	// The option to execute the forward pass for inference only.
	ExecutionOptionsForwardForInference ExecutionOptions = 8
	// The option to enable additional per-layer profiling information using signposts.
	ExecutionOptionsPerLayerProfiling ExecutionOptions = 16
)

func (ExecutionOptions) String added in v0.17.0 

func (e ExecutionOptions) String() string

String returns the ExecutionOptions constant's name, or its numeric form when the value is not a known constant.

type FilesecProperty added in v0.17.0 

type FilesecProperty int32
const (
	FilesecPropertyOwner        FilesecProperty = 1
	FilesecPropertyGroup        FilesecProperty = 2
	FilesecPropertyUUID         FilesecProperty = 3
	FilesecPropertyMode         FilesecProperty = 4
	FilesecPropertyACL          FilesecProperty = 5
	FilesecPropertyGrpuuid      FilesecProperty = 6
	FilesecPropertyACLRaw       FilesecProperty = 100
	FilesecPropertyACLAllocsize FilesecProperty = 101
)

func (FilesecProperty) String added in v0.17.0 

func (e FilesecProperty) String() string

String returns the FilesecProperty constant's name, or its numeric form when the value is not a known constant.

type Flag added in v0.17.0 

type Flag int32
const (
	FlagFlagDeferInherit      Flag = 1
	FlagFlagNoInherit         Flag = 131072
	FlagEntryInherited        Flag = 16
	FlagEntryFileInherit      Flag = 32
	FlagEntryDirectoryInherit Flag = 64
	FlagEntryLimitInherit     Flag = 128
	FlagEntryOnlyInherit      Flag = 256
)

func (Flag) String added in v0.17.0 

func (e Flag) String() string

String returns the Flag constant's name, or its numeric form when the value is not a known constant.

type FullyConnectedLayer added in v0.17.0 

type FullyConnectedLayer struct {
	Layer
}

FullyConnectedLayer is an idiomatic wrapper over the Objective-C class MLCFullyConnectedLayer.

It embeds Layer, promoting that type's methods.

A layer that connects each input to each output within its layer.

func FullyConnectedLayerFromID added in v0.17.0 

func FullyConnectedLayerFromID(id objc.ID) *FullyConnectedLayer

FullyConnectedLayerFromID adopts an existing Objective-C object as a FullyConnectedLayer (nil for 0), retaining it and registering a release finalizer.

func MLCFullyConnectedLayerLayerWithWeightsBiasesDescriptor 

func MLCFullyConnectedLayerLayerWithWeightsBiasesDescriptor(weights *Tensor, biases *Tensor, descriptor *ConvolutionDescriptor) *FullyConnectedLayer

MLCFullyConnectedLayerLayerWithWeightsBiasesDescriptor creates a fully connected layer with the weights, biases, and convolution descriptor you specify.

func NewFullyConnectedLayer added in v0.17.0 

func NewFullyConnectedLayer() *FullyConnectedLayer

NewFullyConnectedLayer creates a new FullyConnectedLayer.

func (*FullyConnectedLayer) Biases added in v0.17.0 

func (fcl *FullyConnectedLayer) Biases() *Tensor

Biases returns the bias tensor used by the convolution layer

func (*FullyConnectedLayer) BiasesParameter added in v0.17.0 

func (fcl *FullyConnectedLayer) BiasesParameter() *TensorParameter

BiasesParameter returns the bias tensor parameter used for optimizer update

func (*FullyConnectedLayer) Descriptor added in v0.17.0 

func (fcl *FullyConnectedLayer) Descriptor() *ConvolutionDescriptor

Descriptor returns the convolution descriptor

func (*FullyConnectedLayer) Weights added in v0.17.0 

func (fcl *FullyConnectedLayer) Weights() *Tensor

Weights returns the weights tensor used by the convolution layer

func (*FullyConnectedLayer) WeightsParameter added in v0.17.0 

func (fcl *FullyConnectedLayer) WeightsParameter() *TensorParameter

WeightsParameter returns the weights tensor parameter used for optimizer update

func (*FullyConnectedLayer) WithIsDebuggingEnabled added in v0.17.0 

func (fcl *FullyConnectedLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *FullyConnectedLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*FullyConnectedLayer) WithLabel added in v0.17.0 

func (fcl *FullyConnectedLayer) WithLabel(label string) *FullyConnectedLayer

WithLabel sets a string that helps identify this layer.

type GatherLayer added in v0.17.0 

type GatherLayer struct {
	Layer
}

GatherLayer is an idiomatic wrapper over the Objective-C class MLCGatherLayer.

It embeds Layer, promoting that type's methods.

A layer that fetches data at the locations you specify.

func GatherLayerFromID added in v0.17.0 

func GatherLayerFromID(id objc.ID) *GatherLayer

GatherLayerFromID adopts an existing Objective-C object as a GatherLayer (nil for 0), retaining it and registering a release finalizer.

func MLCGatherLayerLayerWithDimension 

func MLCGatherLayerLayerWithDimension(dimension int) *GatherLayer

MLCGatherLayerLayerWithDimension creates a gather layer with the dimension you specify.

func NewGatherLayer added in v0.17.0 

func NewGatherLayer() *GatherLayer

NewGatherLayer creates a new GatherLayer.

func (*GatherLayer) Dimension added in v0.17.0 

func (gl *GatherLayer) Dimension() int

Dimension returns the dimension along which to index

func (*GatherLayer) WithIsDebuggingEnabled added in v0.17.0 

func (gl *GatherLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *GatherLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*GatherLayer) WithLabel added in v0.17.0 

func (gl *GatherLayer) WithLabel(label string) *GatherLayer

WithLabel sets a string that helps identify this layer.

type GradientClippingType added in v0.17.0 

type GradientClippingType int32

A clipping type the system applies to a gradient. 

const (
	// An option that clips by value.
	GradientClippingTypeByValue GradientClippingType = 0
	// An option that clips by norm.
	GradientClippingTypeByNorm GradientClippingType = 1
	// An option that clips by global norm.
	GradientClippingTypeByGlobalNorm GradientClippingType = 2
)

func (GradientClippingType) String added in v0.17.0 

func (e GradientClippingType) String() string

String returns the GradientClippingType constant's name, or its numeric form when the value is not a known constant.

type GramMatrixLayer added in v0.17.0 

type GramMatrixLayer struct {
	Layer
}

GramMatrixLayer is an idiomatic wrapper over the Objective-C class MLCGramMatrixLayer.

It embeds Layer, promoting that type's methods.

A layer that computes the uncentered cross-correlation values between the spacial planes of each feature channel of a tensor.

func GramMatrixLayerFromID added in v0.17.0 

func GramMatrixLayerFromID(id objc.ID) *GramMatrixLayer

GramMatrixLayerFromID adopts an existing Objective-C object as a GramMatrixLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithScale added in v0.17.0 

func LayerWithScale(scale float32) *GramMatrixLayer

LayerWithScale creates a gram matrix layer with the scaling factor you specify.

func NewGramMatrixLayer added in v0.17.0 

func NewGramMatrixLayer() *GramMatrixLayer

NewGramMatrixLayer creates a new GramMatrixLayer.

func (*GramMatrixLayer) Scale added in v0.17.0 

func (gml *GramMatrixLayer) Scale() float32

Scale returns the scale factor

func (*GramMatrixLayer) WithIsDebuggingEnabled added in v0.17.0 

func (gml *GramMatrixLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *GramMatrixLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*GramMatrixLayer) WithLabel added in v0.17.0 

func (gml *GramMatrixLayer) WithLabel(label string) *GramMatrixLayer

WithLabel sets a string that helps identify this layer.

type Graph added in v0.17.0 

type Graph struct {
	objref.Handle
}

Graph is an idiomatic wrapper over the Objective-C class MLCGraph.

Graph is an abstract base — you do not construct it directly. Construct one of InferenceGraph, TrainingGraph and pass it where a Graph is accepted.

A graph of layers you use to build a training or inference graph.

func GraphFromID added in v0.17.0 

func GraphFromID(id objc.ID) *Graph

GraphFromID adopts an existing Objective-C object as a Graph (nil for 0), retaining it and registering a release finalizer.

func MLCGraphGraph 

func MLCGraphGraph() *Graph

MLCGraphGraph creates a new graph.

func (*Graph) BindAndWriteDataForInputsToDeviceBatchSizeSynchronous added in v0.17.0 

func (g *Graph) BindAndWriteDataForInputsToDeviceBatchSizeSynchronous(inputsData map[string]*TensorData, inputTensors map[string]*Tensor, device *Device, batchSize int, synchronous bool) bool

BindAndWriteDataForInputsToDeviceBatchSizeSynchronous associates the given data with the input tensors, and if the device is a GPU, also copies the data to the device memory.

func (*Graph) BindAndWriteDataForInputsToDeviceSynchronous added in v0.17.0 

func (g *Graph) BindAndWriteDataForInputsToDeviceSynchronous(inputsData map[string]*TensorData, inputTensors map[string]*Tensor, device *Device, synchronous bool) bool

BindAndWriteDataForInputsToDeviceSynchronous associates the given data with the input tensors, and if the device is a GPU, also copies the data to the device memory.

func (*Graph) ConcatenateWithSourcesDimension added in v0.17.0 

func (g *Graph) ConcatenateWithSourcesDimension(sources []*Tensor, dimension int) *Tensor

ConcatenateWithSourcesDimension adds a new concatenation layer to the graph using the source tensors and concatenation dimension you specify.

func (*Graph) Description added in v0.17.0 

func (g *Graph) Description() string

Description returns the object's -description text.

func (*Graph) Device added in v0.17.0 

func (g *Graph) Device() *Device

Device returns the device to be used when compiling and executing a graph

func (*Graph) GatherWithDimensionSourceIndices added in v0.17.0 

func (g *Graph) GatherWithDimensionSourceIndices(dimension int, source *Tensor, indices *Tensor) *Tensor

GatherWithDimensionSourceIndices adds a gather layer to the graph using the source tensor, dimension along which to index, and the indices you specify.

func (*Graph) IsEqual added in v0.17.0 

func (g *Graph) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*Graph) IsKind added in v0.17.0 

func (g *Graph) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*Graph) Layers added in v0.17.0 

func (g *Graph) Layers() []*Layer

Layers returns layers in the graph

Layers returns the collection as a Go slice.

func (*Graph) NodeWithLayerSource added in v0.17.0 

func (g *Graph) NodeWithLayerSource(layer *Layer, source *Tensor) *Tensor

NodeWithLayerSource adds the layer and source tensor that you specify to the graph.

func (*Graph) NodeWithLayerSources added in v0.17.0 

func (g *Graph) NodeWithLayerSources(layer *Layer, sources []*Tensor) *Tensor

NodeWithLayerSources adds the layer and source tensors that you specify to the graph.

func (*Graph) NodeWithLayerSourcesDisableUpdate added in v0.17.0 

func (g *Graph) NodeWithLayerSourcesDisableUpdate(layer *Layer, sources []*Tensor, disableUpdate bool) *Tensor

NodeWithLayerSourcesDisableUpdate adds the layer, source tensors, and option to disable optimizer updates that you specify to the graph.

func (*Graph) NodeWithLayerSourcesLossLabels added in v0.17.0 

func (g *Graph) NodeWithLayerSourcesLossLabels(layer *Layer, sources []*Tensor, lossLabels []*Tensor) *Tensor

NodeWithLayerSourcesLossLabels adds the layer, sources, and loss labels tensors that you specify to the graph.

func (*Graph) ReshapeWithShapeSource added in v0.17.0 

func (g *Graph) ReshapeWithShapeSource(shape []*foundation.Number, source *Tensor) *Tensor

ReshapeWithShapeSource adds a new reshape layer to the graph using the shape and source tensor you specify.

func (*Graph) ResultTensorsForLayer added in v0.17.0 

func (g *Graph) ResultTensorsForLayer(layer *Layer) []*Tensor

ResultTensorsForLayer gets the result tensors for a layer in the training graph.

func (*Graph) ScatterWithDimensionSourceIndicesCopyFromReductionType added in v0.17.0 

func (g *Graph) ScatterWithDimensionSourceIndicesCopyFromReductionType(dimension int, source *Tensor, indices *Tensor, copyFrom *Tensor, reductionType ReductionType) *Tensor

ScatterWithDimensionSourceIndicesCopyFromReductionType adds a scatter layer to the graph.

func (*Graph) SelectWithSourcesCondition added in v0.17.0 

func (g *Graph) SelectWithSourcesCondition(sources []*Tensor, condition *Tensor) *Tensor

SelectWithSourcesCondition adds a select layer to the graph using the condition mask and source tensors you specify.

func (*Graph) SourceTensorsForLayer added in v0.17.0 

func (g *Graph) SourceTensorsForLayer(layer *Layer) []*Tensor

SourceTensorsForLayer gets the source tensors for a layer in the training graph.

func (*Graph) SplitWithSourceSplitCountDimension added in v0.17.0 

func (g *Graph) SplitWithSourceSplitCountDimension(source *Tensor, splitCount int, dimension int) []*Tensor

SplitWithSourceSplitCountDimension adds a new split layer to the graph using the source tensor, number of splits, and dimension to split the source tensor that you specify.

func (*Graph) SplitWithSourceSplitSectionLengthsDimension added in v0.17.0 

func (g *Graph) SplitWithSourceSplitSectionLengthsDimension(source *Tensor, splitSectionLengths []*foundation.Number, dimension int) []*Tensor

SplitWithSourceSplitSectionLengthsDimension adds a new split layer to the graph using the source tensor, lengths of each split section, and dimension to split the source tensor that you specify.

func (*Graph) String added in v0.17.0 

func (g *Graph) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*Graph) SummarizedDOTDescription added in v0.17.0 

func (g *Graph) SummarizedDOTDescription() string

SummarizedDOTDescription returns a DOT representation of the graph. For more info on the DOT language, refer to https://en.wikipedia.org/wiki/DOT_(graph_description_language). Edges that have a dashed lines are those that have stop gradients, while those with solid lines don't.

func (*Graph) TransposeWithDimensionsSource added in v0.17.0 

func (g *Graph) TransposeWithDimensionsSource(dimensions []*foundation.Number, source *Tensor) *Tensor

TransposeWithDimensionsSource adds a new transpose layer to the graph using the dimensions and source tensor you specify.

type GraphCompilationOptions added in v0.17.0 

type GraphCompilationOptions uint64

A bitmask that specifies the options you use when compiling a graph. Bitmask — values may be combined with |. 

const (
	// The default option for graph compilation.
	GraphCompilationOptionsNone GraphCompilationOptions = 0
	// The option to debug layers during graph compilation.
	GraphCompilationOptionsDebugLayers GraphCompilationOptions = 1
	// The option to disable layer fusion during graph compilation.
	GraphCompilationOptionsDisableLayerFusion GraphCompilationOptions = 2
	// The option to link graphs during graph compilation.
	GraphCompilationOptionsLinkGraphs GraphCompilationOptions = 4
	// The option to compute all gradients during graph compilation.
	GraphCompilationOptionsComputeAllGradients GraphCompilationOptions = 8
)

func (GraphCompilationOptions) String added in v0.17.0 

func (e GraphCompilationOptions) String() string

String returns the GraphCompilationOptions constant's name, or its numeric form when the value is not a known constant.

type GraphProvider added in v0.17.0 

type GraphProvider interface {
	objref.Object
	// contains filtered or unexported methods
}

GraphProvider is accepted wherever a MLCGraph (or one of its subclasses) is expected.

type GroupNormalizationLayer added in v0.17.0 

type GroupNormalizationLayer struct {
	Layer
}

GroupNormalizationLayer is an idiomatic wrapper over the Objective-C class MLCGroupNormalizationLayer.

It embeds Layer, promoting that type's methods.

A layer that divides the channels into groups for normalization.

func GroupNormalizationLayerFromID added in v0.17.0 

func GroupNormalizationLayerFromID(id objc.ID) *GroupNormalizationLayer

GroupNormalizationLayerFromID adopts an existing Objective-C object as a GroupNormalizationLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithFeatureChannelCountGroupCountBetaGammaVarianceEpsilon added in v0.17.0 

func LayerWithFeatureChannelCountGroupCountBetaGammaVarianceEpsilon(featureChannelCount int, groupCount int, beta *Tensor, gamma *Tensor, varianceEpsilon float32) *GroupNormalizationLayer

LayerWithFeatureChannelCountGroupCountBetaGammaVarianceEpsilon creates a group normalization layer with the number of feature channels and groups, beta and gamma tensors, and variance epsilon you specify.

func NewGroupNormalizationLayer added in v0.17.0 

func NewGroupNormalizationLayer() *GroupNormalizationLayer

NewGroupNormalizationLayer creates a new GroupNormalizationLayer.

func (*GroupNormalizationLayer) Beta added in v0.17.0 

func (gnl *GroupNormalizationLayer) Beta() *Tensor

Beta returns the beta tensor

func (*GroupNormalizationLayer) BetaParameter added in v0.17.0 

func (gnl *GroupNormalizationLayer) BetaParameter() *TensorParameter

BetaParameter returns the beta tensor parameter used for optimizer update

func (*GroupNormalizationLayer) FeatureChannelCount added in v0.17.0 

func (gnl *GroupNormalizationLayer) FeatureChannelCount() int

FeatureChannelCount returns the number of feature channels

func (*GroupNormalizationLayer) Gamma added in v0.17.0 

func (gnl *GroupNormalizationLayer) Gamma() *Tensor

Gamma returns the gamma tensor

func (*GroupNormalizationLayer) GammaParameter added in v0.17.0 

func (gnl *GroupNormalizationLayer) GammaParameter() *TensorParameter

GammaParameter returns the gamma tensor parameter used for optimizer update

func (*GroupNormalizationLayer) GroupCount added in v0.17.0 

func (gnl *GroupNormalizationLayer) GroupCount() int

GroupCount returns the number of groups to separate the channels into

func (*GroupNormalizationLayer) VarianceEpsilon added in v0.17.0 

func (gnl *GroupNormalizationLayer) VarianceEpsilon() float32

VarianceEpsilon returns a value used for numerical stability

func (*GroupNormalizationLayer) WithIsDebuggingEnabled added in v0.17.0 

func (gnl *GroupNormalizationLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *GroupNormalizationLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*GroupNormalizationLayer) WithLabel added in v0.17.0 

func (gnl *GroupNormalizationLayer) WithLabel(label string) *GroupNormalizationLayer

WithLabel sets a string that helps identify this layer.

type Idtype added in v0.17.0 

type Idtype int32
const (
	IdtypeAll  Idtype = 0
	IdtypePid  Idtype = 1
	IdtypePgid Idtype = 2
)

func (Idtype) String added in v0.17.0 

func (e Idtype) String() string

String returns the Idtype constant's name, or its numeric form when the value is not a known constant.

type InferenceGraph added in v0.17.0 

type InferenceGraph struct {
	Graph
}

InferenceGraph is an idiomatic wrapper over the Objective-C class MLCInferenceGraph.

It embeds Graph, promoting that type's methods.

An inference graph created from one or more MLCGraph instances plus additional layers added directly to the inference graph.

func GraphWithGraphObjects added in v0.17.0 

func GraphWithGraphObjects(graphObjects []*Graph) *InferenceGraph

GraphWithGraphObjects creates an inference graph with the layers from the graph objects you specify.

func InferenceGraphFromID added in v0.17.0 

func InferenceGraphFromID(id objc.ID) *InferenceGraph

InferenceGraphFromID adopts an existing Objective-C object as a InferenceGraph (nil for 0), retaining it and registering a release finalizer.

func NewInferenceGraph added in v0.17.0 

func NewInferenceGraph() *InferenceGraph

NewInferenceGraph creates a new InferenceGraph.

func (*InferenceGraph) AddInputs added in v0.17.0 

func (ig *InferenceGraph) AddInputs(inputs map[string]*Tensor) bool

AddInputs adds the inputs you specify to the inference graph.

func (*InferenceGraph) AddInputsLossLabelsLossLabelWeights added in v0.17.0 

func (ig *InferenceGraph) AddInputsLossLabelsLossLabelWeights(inputs map[string]*Tensor, lossLabels map[string]*Tensor, lossLabelWeights map[string]*Tensor) bool

AddInputsLossLabelsLossLabelWeights adds the inputs, loss labels, and loss label weights that you specify to the inference graph.

func (*InferenceGraph) AddOutputs added in v0.17.0 

func (ig *InferenceGraph) AddOutputs(outputs map[string]*Tensor) bool

AddOutputs adds the outputs you specify to the inference graph.

func (*InferenceGraph) CompileWithOptionsDevice added in v0.17.0 

func (ig *InferenceGraph) CompileWithOptionsDevice(options GraphCompilationOptions, device *Device) bool

CompileWithOptionsDevice compiles the inference graph for the options and device you specify.

func (*InferenceGraph) CompileWithOptionsDeviceInputTensorsInputTensorsData added in v0.17.0 

func (ig *InferenceGraph) CompileWithOptionsDeviceInputTensorsInputTensorsData(options GraphCompilationOptions, device *Device, inputTensors map[string]*Tensor, inputTensorsData map[string]*TensorData) bool

CompileWithOptionsDeviceInputTensorsInputTensorsData compiles the inference graph for the options, device, and input tensors you specify.

func (*InferenceGraph) DeviceMemorySize added in v0.17.0 

func (ig *InferenceGraph) DeviceMemorySize() int

DeviceMemorySize returns the total size in bytes of device memory used by all intermediate tensors in the inference graph

func (*InferenceGraph) ExecuteWithInputsDataBatchSizeOptionsCompletionHandler added in v0.18.0 

func (ig *InferenceGraph) ExecuteWithInputsDataBatchSizeOptionsCompletionHandler(inputsData map[string]*TensorData, batchSize int, options ExecutionOptions, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteWithInputsDataBatchSizeOptionsCompletionHandler executes the inference graph with the inputs data, batch size, execution options, and completion handler you specify.

func (*InferenceGraph) ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataBatchSizeOptionsCompletionHandler added in v0.18.0 

func (ig *InferenceGraph) ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataBatchSizeOptionsCompletionHandler(inputsData map[string]*TensorData, lossLabelsData map[string]*TensorData, lossLabelWeightsData map[string]*TensorData, batchSize int, options ExecutionOptions, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataBatchSizeOptionsCompletionHandler executes the inference graph with the input data, batch size, execution options and completion handler you specify.

func (*InferenceGraph) ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataOutputsDataBatchSizeOptionsCompletionHandler added in v0.18.0 

func (ig *InferenceGraph) ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataOutputsDataBatchSizeOptionsCompletionHandler(inputsData map[string]*TensorData, lossLabelsData map[string]*TensorData, lossLabelWeightsData map[string]*TensorData, outputsData map[string]*TensorData, batchSize int, options ExecutionOptions, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataOutputsDataBatchSizeOptionsCompletionHandler executes the inference graph with the input and output data, batch size, execution options, and completion handler that you specify.

func (*InferenceGraph) ExecuteWithInputsDataOutputsDataBatchSizeOptionsCompletionHandler added in v0.18.0 

func (ig *InferenceGraph) ExecuteWithInputsDataOutputsDataBatchSizeOptionsCompletionHandler(inputsData map[string]*TensorData, outputsData map[string]*TensorData, batchSize int, options ExecutionOptions, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteWithInputsDataOutputsDataBatchSizeOptionsCompletionHandler executes the inference graph with the inputs and outputs data, batch size, execution options, and completion handler that you specify.

func (*InferenceGraph) LinkWithGraphs added in v0.17.0 

func (ig *InferenceGraph) LinkWithGraphs(graphs []*InferenceGraph) bool

LinkWithGraphs links the inference graphs you specify.

type InstanceNormalizationLayer added in v0.17.0 

type InstanceNormalizationLayer struct {
	Layer
}

InstanceNormalizationLayer is an idiomatic wrapper over the Objective-C class MLCInstanceNormalizationLayer.

It embeds Layer, promoting that type's methods.

A layer that normalizes all features of one channel.

func InstanceNormalizationLayerFromID added in v0.17.0 

func InstanceNormalizationLayerFromID(id objc.ID) *InstanceNormalizationLayer

InstanceNormalizationLayerFromID adopts an existing Objective-C object as a InstanceNormalizationLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithFeatureChannelCountBetaGammaVarianceEpsilon added in v0.17.0 

func LayerWithFeatureChannelCountBetaGammaVarianceEpsilon(featureChannelCount int, beta *Tensor, gamma *Tensor, varianceEpsilon float32) *InstanceNormalizationLayer

LayerWithFeatureChannelCountBetaGammaVarianceEpsilon creates an instance normalization layer with the number of feature channels, beta and gamma tensors, and variance epsilon you specify.

func LayerWithFeatureChannelCountBetaGammaVarianceEpsilonMomentum added in v0.17.0 

func LayerWithFeatureChannelCountBetaGammaVarianceEpsilonMomentum(featureChannelCount int, beta *Tensor, gamma *Tensor, varianceEpsilon float32, momentum float32) *InstanceNormalizationLayer

LayerWithFeatureChannelCountBetaGammaVarianceEpsilonMomentum creates an instance normalization layer with the number of feature channels, beta and gamma tensors, variance epsilon, and momentum you specify.

func MLCInstanceNormalizationLayerLayerWithFeatureChannelCountMeanVarianceBetaGammaVarianceEpsilonMomentum 

func MLCInstanceNormalizationLayerLayerWithFeatureChannelCountMeanVarianceBetaGammaVarianceEpsilonMomentum(featureChannelCount int, mean *Tensor, variance *Tensor, beta *Tensor, gamma *Tensor, varianceEpsilon float32, momentum float32) *InstanceNormalizationLayer

MLCInstanceNormalizationLayerLayerWithFeatureChannelCountMeanVarianceBetaGammaVarianceEpsilonMomentum creates an instance normalization layer with the number of feature channels, mean, variance, beta and gamma tensors, variance epsilon, and momentum you specify.

func NewInstanceNormalizationLayer added in v0.17.0 

func NewInstanceNormalizationLayer() *InstanceNormalizationLayer

NewInstanceNormalizationLayer creates a new InstanceNormalizationLayer.

func (*InstanceNormalizationLayer) Beta added in v0.17.0 

func (inl *InstanceNormalizationLayer) Beta() *Tensor

Beta returns the beta tensor

func (*InstanceNormalizationLayer) BetaParameter added in v0.17.0 

func (inl *InstanceNormalizationLayer) BetaParameter() *TensorParameter

BetaParameter returns the beta tensor parameter used for optimizer update

func (*InstanceNormalizationLayer) FeatureChannelCount added in v0.17.0 

func (inl *InstanceNormalizationLayer) FeatureChannelCount() int

FeatureChannelCount returns the number of feature channels

func (*InstanceNormalizationLayer) Gamma added in v0.17.0 

func (inl *InstanceNormalizationLayer) Gamma() *Tensor

Gamma returns the gamma tensor

func (*InstanceNormalizationLayer) GammaParameter added in v0.17.0 

func (inl *InstanceNormalizationLayer) GammaParameter() *TensorParameter

GammaParameter returns the gamma tensor parameter used for optimizer update

func (*InstanceNormalizationLayer) Mean added in v0.17.0 

func (inl *InstanceNormalizationLayer) Mean() *Tensor

Mean returns the running mean tensor

func (*InstanceNormalizationLayer) Momentum added in v0.17.0 

func (inl *InstanceNormalizationLayer) Momentum() float32

Momentum returns the value used for the running mean and variance computation The default is 0.99f.

func (*InstanceNormalizationLayer) Variance added in v0.17.0 

func (inl *InstanceNormalizationLayer) Variance() *Tensor

Variance returns the running variance tensor

func (*InstanceNormalizationLayer) VarianceEpsilon added in v0.17.0 

func (inl *InstanceNormalizationLayer) VarianceEpsilon() float32

VarianceEpsilon returns a value used for numerical stability

func (*InstanceNormalizationLayer) WithIsDebuggingEnabled added in v0.17.0 

func (inl *InstanceNormalizationLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *InstanceNormalizationLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*InstanceNormalizationLayer) WithLabel added in v0.17.0 

func (inl *InstanceNormalizationLayer) WithLabel(label string) *InstanceNormalizationLayer

WithLabel sets a string that helps identify this layer.

type IpcInfoObjectType added in v0.17.0 

type IpcInfoObjectType uint32
const (
	IpcInfoObjectTypeNone               IpcInfoObjectType = 0
	IpcInfoObjectTypeThreadControl      IpcInfoObjectType = 1
	IpcInfoObjectTypeTaskControl        IpcInfoObjectType = 2
	IpcInfoObjectTypeHost               IpcInfoObjectType = 3
	IpcInfoObjectTypeHostPriv           IpcInfoObjectType = 4
	IpcInfoObjectTypeProcessor          IpcInfoObjectType = 5
	IpcInfoObjectTypeProcessorSet       IpcInfoObjectType = 6
	IpcInfoObjectTypeProcessorSetName   IpcInfoObjectType = 7
	IpcInfoObjectTypeTimer              IpcInfoObjectType = 8
	IpcInfoObjectTypePortSubstOnce      IpcInfoObjectType = 9
	IpcInfoObjectTypeMig                IpcInfoObjectType = 10
	IpcInfoObjectTypeMemoryObject       IpcInfoObjectType = 11
	IpcInfoObjectTypeXmmPager           IpcInfoObjectType = 12
	IpcInfoObjectTypeXmmKernel          IpcInfoObjectType = 13
	IpcInfoObjectTypeXmmReply           IpcInfoObjectType = 14
	IpcInfoObjectTypeUndReply           IpcInfoObjectType = 15
	IpcInfoObjectTypeHostNotify         IpcInfoObjectType = 16
	IpcInfoObjectTypeHostSecurity       IpcInfoObjectType = 17
	IpcInfoObjectTypeLedger             IpcInfoObjectType = 18
	IpcInfoObjectTypeMainDevice         IpcInfoObjectType = 19
	IpcInfoObjectTypeTaskName           IpcInfoObjectType = 20
	IpcInfoObjectTypeSubsystem          IpcInfoObjectType = 21
	IpcInfoObjectTypeIODoneQueue        IpcInfoObjectType = 22
	IpcInfoObjectTypeSemaphore          IpcInfoObjectType = 23
	IpcInfoObjectTypeLockSet            IpcInfoObjectType = 24
	IpcInfoObjectTypeClock              IpcInfoObjectType = 25
	IpcInfoObjectTypeClockCtrl          IpcInfoObjectType = 26
	IpcInfoObjectTypeIokitIdent         IpcInfoObjectType = 27
	IpcInfoObjectTypeNamedEntry         IpcInfoObjectType = 28
	IpcInfoObjectTypeIokitConnect       IpcInfoObjectType = 29
	IpcInfoObjectTypeIokitObject        IpcInfoObjectType = 30
	IpcInfoObjectTypeUpl                IpcInfoObjectType = 31
	IpcInfoObjectTypeMemObjControl      IpcInfoObjectType = 32
	IpcInfoObjectTypeAuSessionport      IpcInfoObjectType = 33
	IpcInfoObjectTypeFileport           IpcInfoObjectType = 34
	IpcInfoObjectTypeLabelh             IpcInfoObjectType = 35
	IpcInfoObjectTypeTaskResume         IpcInfoObjectType = 36
	IpcInfoObjectTypeVoucher            IpcInfoObjectType = 37
	IpcInfoObjectTypeVoucherAttrControl IpcInfoObjectType = 38
	IpcInfoObjectTypeWorkInterval       IpcInfoObjectType = 39
	IpcInfoObjectTypeUxHandler          IpcInfoObjectType = 40
	IpcInfoObjectTypeUextObject         IpcInfoObjectType = 41
	IpcInfoObjectTypeArcadeReg          IpcInfoObjectType = 42
	IpcInfoObjectTypeEventlink          IpcInfoObjectType = 43
	IpcInfoObjectTypeTaskInspect        IpcInfoObjectType = 44
	IpcInfoObjectTypeTaskRead           IpcInfoObjectType = 45
	IpcInfoObjectTypeThreadInspect      IpcInfoObjectType = 46
	IpcInfoObjectTypeThreadRead         IpcInfoObjectType = 47
	IpcInfoObjectTypeSuidCred           IpcInfoObjectType = 48
	IpcInfoObjectTypeHypervisor         IpcInfoObjectType = 49
	IpcInfoObjectTypeTaskIDToken        IpcInfoObjectType = 50
	IpcInfoObjectTypeTaskFatal          IpcInfoObjectType = 51
	IpcInfoObjectTypeKcdata             IpcInfoObjectType = 52
	IpcInfoObjectTypeExclavesResource   IpcInfoObjectType = 53
	IpcInfoObjectTypeThreadResume       IpcInfoObjectType = 54
	IpcInfoObjectTypeUnknown            IpcInfoObjectType = 4294967295
)

func (IpcInfoObjectType) String added in v0.17.0 

func (e IpcInfoObjectType) String() string

String returns the IpcInfoObjectType constant's name, or its numeric form when the value is not a known constant.

type LSTMDescriptor added in v0.17.0 

type LSTMDescriptor struct {
	objref.Handle
}

LSTMDescriptor is an idiomatic wrapper over the Objective-C class MLCLSTMDescriptor.

The configuration object you use to create the LSTM layer.

func DescriptorWithInputSizeHiddenSizeLayerCount added in v0.17.0 

func DescriptorWithInputSizeHiddenSizeLayerCount(inputSize int, hiddenSize int, layerCount int) *LSTMDescriptor

DescriptorWithInputSizeHiddenSizeLayerCount creates a batch first LSTM descriptor with the input size and number of layers you specify.

func DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesBatchFirstIsBidirectionalDropout added in v0.17.0 

func DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesBatchFirstIsBidirectionalDropout(inputSize int, hiddenSize int, layerCount int, usesBiases bool, batchFirst bool, isBidirectional bool, dropout float32) *LSTMDescriptor

DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesBatchFirstIsBidirectionalDropout creates a batch first LSTM descriptor that allows you to indicate whether the input and output shape is batch first.

func DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesBatchFirstIsBidirectionalReturnsSequencesDropout added in v0.17.0 

func DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesBatchFirstIsBidirectionalReturnsSequencesDropout(inputSize int, hiddenSize int, layerCount int, usesBiases bool, batchFirst bool, isBidirectional bool, returnsSequences bool, dropout float32) *LSTMDescriptor

DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesBatchFirstIsBidirectionalReturnsSequencesDropout creates a batch first LSTM descriptor that allows you to indicate whether the layer returns output for all sequences, or output for only the last sequence.

func DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesBatchFirstIsBidirectionalReturnsSequencesDropoutResultMode added in v0.17.0 

func DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesBatchFirstIsBidirectionalReturnsSequencesDropoutResultMode(inputSize int, hiddenSize int, layerCount int, usesBiases bool, batchFirst bool, isBidirectional bool, returnsSequences bool, dropout float32, resultMode LSTMResultMode) *LSTMDescriptor

DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesBatchFirstIsBidirectionalReturnsSequencesDropoutResultMode creates a descriptor with the number of features and layers, dropout, and options for use of biases, batch order, return sequences, bidirectionality, and expected tensors you specify.

func DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesIsBidirectionalDropout added in v0.17.0 

func DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesIsBidirectionalDropout(inputSize int, hiddenSize int, layerCount int, usesBiases bool, isBidirectional bool, dropout float32) *LSTMDescriptor

DescriptorWithInputSizeHiddenSizeLayerCountUsesBiasesIsBidirectionalDropout creates a batch first LSTM descriptor with bias and bidirectional options you specify.

func LSTMDescriptorFromID added in v0.17.0 

func LSTMDescriptorFromID(id objc.ID) *LSTMDescriptor

LSTMDescriptorFromID adopts an existing Objective-C object as a LSTMDescriptor (nil for 0), retaining it and registering a release finalizer.

func NewLSTMDescriptor added in v0.17.0 

func NewLSTMDescriptor() *LSTMDescriptor

NewLSTMDescriptor creates a new LSTMDescriptor.

func (*LSTMDescriptor) BatchFirst added in v0.17.0 

func (ld *LSTMDescriptor) BatchFirst() bool

BatchFirst reports whether LSTM only supports batchFirst=YES. This means the input and output will have shape [batch size, time steps, feature]. Default is true.

func (*LSTMDescriptor) Description added in v0.17.0 

func (ld *LSTMDescriptor) Description() string

Description returns the object's -description text.

func (*LSTMDescriptor) Dropout added in v0.17.0 

func (ld *LSTMDescriptor) Dropout() float32

Dropout returns if non-zero, intrdouces a dropout layer on the outputs of each LSTM layer except the last layer, with dropout probablity equal to dropout. Default is 0.0.

func (*LSTMDescriptor) HiddenSize added in v0.17.0 

func (ld *LSTMDescriptor) HiddenSize() int

HiddenSize returns the number of feature channels in the hidden state

func (*LSTMDescriptor) InputSize added in v0.17.0 

func (ld *LSTMDescriptor) InputSize() int

InputSize returns the number of expected feature channels in the input

func (*LSTMDescriptor) IsBidirectional added in v0.17.0 

func (ld *LSTMDescriptor) IsBidirectional() bool

IsBidirectional reports whether if true, becomes a bidirectional LSTM. Default is false.

func (*LSTMDescriptor) IsEqual added in v0.17.0 

func (ld *LSTMDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*LSTMDescriptor) IsKind added in v0.17.0 

func (ld *LSTMDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*LSTMDescriptor) LayerCount added in v0.17.0 

func (ld *LSTMDescriptor) LayerCount() int

LayerCount returns the number of recurrent layers. Default is 1.

func (*LSTMDescriptor) ResultMode added in v0.17.0 

func (ld *LSTMDescriptor) ResultMode() LSTMResultMode

ResultMode returns MLCLSTMResultModeOutput returns output data. MLCLSTMResultModeOutputAndStates returns output data, last hidden state h_n, and last cell state c_n. Default MLCLSTMResultModeOutput.

func (*LSTMDescriptor) ReturnsSequences added in v0.17.0 

func (ld *LSTMDescriptor) ReturnsSequences() bool

ReturnsSequences reports whether if true return output for all sequences else return output only for the last sequences. Default: true

func (*LSTMDescriptor) String added in v0.17.0 

func (ld *LSTMDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*LSTMDescriptor) UsesBiases added in v0.17.0 

func (ld *LSTMDescriptor) UsesBiases() bool

UsesBiases reports whether if false, the layer does not use bias terms. Default is true.

type LSTMLayer added in v0.17.0 

type LSTMLayer struct {
	Layer
}

LSTMLayer is an idiomatic wrapper over the Objective-C class MLCLSTMLayer.

It embeds Layer, promoting that type's methods.

A layer that represents long short-term memory (LSTM) networks.

func LSTMLayerFromID added in v0.17.0 

func LSTMLayerFromID(id objc.ID) *LSTMLayer

LSTMLayerFromID adopts an existing Objective-C object as a LSTMLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithDescriptorInputWeightsHiddenWeightsBiases added in v0.17.0 

func LayerWithDescriptorInputWeightsHiddenWeightsBiases(descriptor *LSTMDescriptor, inputWeights []*Tensor, hiddenWeights []*Tensor, biases []*Tensor) *LSTMLayer

LayerWithDescriptorInputWeightsHiddenWeightsBiases creates an LSTM layer with the descriptor, input and hidden weights, and biases you specify.

func LayerWithDescriptorInputWeightsHiddenWeightsPeepholeWeightsBiases added in v0.17.0 

func LayerWithDescriptorInputWeightsHiddenWeightsPeepholeWeightsBiases(descriptor *LSTMDescriptor, inputWeights []*Tensor, hiddenWeights []*Tensor, peepholeWeights []*Tensor, biases []*Tensor) *LSTMLayer

LayerWithDescriptorInputWeightsHiddenWeightsPeepholeWeightsBiases creates an LSTM layer with the descriptor, weights, and biases you specify.

func LayerWithDescriptorInputWeightsHiddenWeightsPeepholeWeightsBiasesGateActivationsOutputResultActivation added in v0.17.0 

func LayerWithDescriptorInputWeightsHiddenWeightsPeepholeWeightsBiasesGateActivationsOutputResultActivation(descriptor *LSTMDescriptor, inputWeights []*Tensor, hiddenWeights []*Tensor, peepholeWeights []*Tensor, biases []*Tensor, gateActivations []*ActivationDescriptor, outputResultActivation *ActivationDescriptor) *LSTMLayer

LayerWithDescriptorInputWeightsHiddenWeightsPeepholeWeightsBiasesGateActivationsOutputResultActivation creates an LSTM layer using the descriptor, weights, biases, gate activations, and output result activation that you specify.

func NewLSTMLayer added in v0.17.0 

func NewLSTMLayer() *LSTMLayer

NewLSTMLayer creates a new LSTMLayer.

func (*LSTMLayer) Biases added in v0.17.0 

func (ll *LSTMLayer) Biases() []*Tensor

Biases returns the array of tensors describing the bias terms for the input, hidden, cell and output gates

Biases returns the collection as a Go slice.

func (*LSTMLayer) BiasesParameters added in v0.17.0 

func (ll *LSTMLayer) BiasesParameters() []*TensorParameter

BiasesParameters returns the bias tensor parameter used for optimizer update

BiasesParameters returns the collection as a Go slice.

func (*LSTMLayer) Descriptor added in v0.17.0 

func (ll *LSTMLayer) Descriptor() *LSTMDescriptor

Descriptor returns the LSTM descriptor

func (*LSTMLayer) GateActivations added in v0.17.0 

func (ll *LSTMLayer) GateActivations() []*ActivationDescriptor

GateActivations returns the array of gate activations for input, hidden, cell and output gates The default gate activations are: sigmoid, sigmoid, tanh, sigmoid

GateActivations returns the collection as a Go slice.

func (*LSTMLayer) HiddenWeights added in v0.17.0 

func (ll *LSTMLayer) HiddenWeights() []*Tensor

HiddenWeights returns the array of tensors describing the hidden weights for the input, hidden, cell and output gates

HiddenWeights returns the collection as a Go slice.

func (*LSTMLayer) HiddenWeightsParameters added in v0.17.0 

func (ll *LSTMLayer) HiddenWeightsParameters() []*TensorParameter

HiddenWeightsParameters returns the hidden weights tensor parameters used for optimizer update

HiddenWeightsParameters returns the collection as a Go slice.

func (*LSTMLayer) InputWeights added in v0.17.0 

func (ll *LSTMLayer) InputWeights() []*Tensor

InputWeights returns the array of tensors describing the input weights for the input, hidden, cell and output gates

InputWeights returns the collection as a Go slice.

func (*LSTMLayer) InputWeightsParameters added in v0.17.0 

func (ll *LSTMLayer) InputWeightsParameters() []*TensorParameter

InputWeightsParameters returns the input weights tensor parameters used for optimizer update

InputWeightsParameters returns the collection as a Go slice.

func (*LSTMLayer) OutputResultActivation added in v0.17.0 

func (ll *LSTMLayer) OutputResultActivation() *ActivationDescriptor

OutputResultActivation returns the output activation descriptor

func (*LSTMLayer) PeepholeWeights added in v0.17.0 

func (ll *LSTMLayer) PeepholeWeights() []*Tensor

PeepholeWeights returns the array of tensors describing the peephole weights for the input, hidden, cell and output gates

PeepholeWeights returns the collection as a Go slice.

func (*LSTMLayer) PeepholeWeightsParameters added in v0.17.0 

func (ll *LSTMLayer) PeepholeWeightsParameters() []*TensorParameter

PeepholeWeightsParameters returns the peephole weights tensor parameters used for optimizer update

PeepholeWeightsParameters returns the collection as a Go slice.

func (*LSTMLayer) WithIsDebuggingEnabled added in v0.17.0 

func (ll *LSTMLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *LSTMLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*LSTMLayer) WithLabel added in v0.17.0 

func (ll *LSTMLayer) WithLabel(label string) *LSTMLayer

WithLabel sets a string that helps identify this layer.

type LSTMResultMode added in v0.17.0 

type LSTMResultMode uint64

Constants that describe the result of an LSTM layer. 

const (
	// A result mode that indicates the layer produces a single result tensor that represents the final output of the LSTM.
	LSTMResultModeOutput LSTMResultMode = 0
	// A result mode that indicates the layer produces three result tensors that represent the final output of the LSTM, the last hidden state, and the cell state.
	LSTMResultModeOutputAndStates LSTMResultMode = 1
)

func (LSTMResultMode) String added in v0.17.0 

func (e LSTMResultMode) String() string

String returns the LSTMResultMode constant's name, or its numeric form when the value is not a known constant.

type LaunchDataType added in v0.17.0 

type LaunchDataType int32
const (
	LaunchDataTypeDictionary LaunchDataType = 1
	LaunchDataTypeArray      LaunchDataType = 2
	LaunchDataTypeFd         LaunchDataType = 3
	LaunchDataTypeInteger    LaunchDataType = 4
	LaunchDataTypeReal       LaunchDataType = 5
	LaunchDataTypeBool       LaunchDataType = 6
	LaunchDataTypeString     LaunchDataType = 7
	LaunchDataTypeOpaque     LaunchDataType = 8
	LaunchDataTypeErrno      LaunchDataType = 9
	LaunchDataTypeMachport   LaunchDataType = 10
)

func (LaunchDataType) String added in v0.17.0 

func (e LaunchDataType) String() string

String returns the LaunchDataType constant's name, or its numeric form when the value is not a known constant.

type Layer added in v0.17.0 

type Layer struct {
	objref.Handle
}

Layer is an idiomatic wrapper over the Objective-C class MLCLayer.

Layer is an abstract base — you do not construct it directly. Construct one of ActivationLayer, ArithmeticLayer, BatchNormalizationLayer, ComparisonLayer, ConcatenationLayer, ConvolutionLayer, DropoutLayer, EmbeddingLayer, FullyConnectedLayer, GatherLayer, GramMatrixLayer, GroupNormalizationLayer, InstanceNormalizationLayer, LSTMLayer, LayerNormalizationLayer, LossLayer, MatMulLayer, MultiheadAttentionLayer, PaddingLayer, PoolingLayer, ReductionLayer, ReshapeLayer, ScatterLayer, SelectionLayer, SliceLayer, SoftmaxLayer, SplitLayer, TransposeLayer, UpsampleLayer and pass it where a Layer is accepted.

The base class for all framework layers.

func LayerFromID added in v0.17.0 

func LayerFromID(id objc.ID) *Layer

LayerFromID adopts an existing Objective-C object as a Layer (nil for 0), retaining it and registering a release finalizer.

func (*Layer) Description added in v0.17.0 

func (l *Layer) Description() string

Description returns the object's -description text.

func (*Layer) DeviceType added in v0.17.0 

func (l *Layer) DeviceType() DeviceType

DeviceType returns the device type where this layer will be executed Typically the MLCDevice passed to compileWithOptions will be the device used to execute layers in the graph. If MLCDeviceTypeANE is selected, it is possible that some of the layers of the graph may not be executed on the ANE but instead on the CPU or GPU. This property can be used to determine which device type the layer will be executed on.

func (*Layer) IsDebuggingEnabled added in v0.17.0 

func (l *Layer) IsDebuggingEnabled() bool

IsDebuggingEnabled reports whether a flag to identify if we want to debug this layer when executing a graph that includes this layer If this is set, we will make sure that the result tensor and gradient tensors are available for reading on CPU The default is false. If isDebuggingEnabled is set to true, make sure to set options to enable debugging when compiling the graph. Otherwise this property may be ignored.

func (*Layer) IsEqual added in v0.17.0 

func (l *Layer) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*Layer) IsKind added in v0.17.0 

func (l *Layer) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*Layer) Label added in v0.17.0 

func (l *Layer) Label() string

Label returns a string to help identify this object.

func (*Layer) LayerID added in v0.17.0 

func (l *Layer) LayerID() int

LayerID returns the layer ID A unique number to identify each layer. Assigned when the layer is created.

func (*Layer) String added in v0.17.0 

func (l *Layer) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*Layer) WithIsDebuggingEnabled added in v0.17.0 

func (l *Layer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *Layer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*Layer) WithLabel added in v0.17.0 

func (l *Layer) WithLabel(label string) *Layer

WithLabel sets a string that helps identify this layer.

type LayerNormalizationLayer added in v0.17.0 

type LayerNormalizationLayer struct {
	Layer
}

LayerNormalizationLayer is an idiomatic wrapper over the Objective-C class MLCLayerNormalizationLayer.

It embeds Layer, promoting that type's methods.

A layer that applies layer normalization over inputs.

func LayerNormalizationLayerFromID added in v0.17.0 

func LayerNormalizationLayerFromID(id objc.ID) *LayerNormalizationLayer

LayerNormalizationLayerFromID adopts an existing Objective-C object as a LayerNormalizationLayer (nil for 0), retaining it and registering a release finalizer.

func LayerWithNormalizedShapeBetaGammaVarianceEpsilon added in v0.17.0 

func LayerWithNormalizedShapeBetaGammaVarianceEpsilon(normalizedShape []*foundation.Number, beta *Tensor, gamma *Tensor, varianceEpsilon float32) *LayerNormalizationLayer

LayerWithNormalizedShapeBetaGammaVarianceEpsilon creates a normalization layer with a shape, beta and gamma tensors, and variance epsilon you specify.

func NewLayerNormalizationLayer added in v0.17.0 

func NewLayerNormalizationLayer() *LayerNormalizationLayer

NewLayerNormalizationLayer creates a new LayerNormalizationLayer.

func (*LayerNormalizationLayer) Beta added in v0.17.0 

func (lnl *LayerNormalizationLayer) Beta() *Tensor

Beta returns the beta tensor

func (*LayerNormalizationLayer) BetaParameter added in v0.17.0 

func (lnl *LayerNormalizationLayer) BetaParameter() *TensorParameter

BetaParameter returns the beta tensor parameter used for optimizer update

func (*LayerNormalizationLayer) Gamma added in v0.17.0 

func (lnl *LayerNormalizationLayer) Gamma() *Tensor

Gamma returns the gamma tensor

func (*LayerNormalizationLayer) GammaParameter added in v0.17.0 

func (lnl *LayerNormalizationLayer) GammaParameter() *TensorParameter

GammaParameter returns the gamma tensor parameter used for optimizer update

func (*LayerNormalizationLayer) NormalizedShape added in v0.17.0 

func (lnl *LayerNormalizationLayer) NormalizedShape() []obj.Object

NormalizedShape returns the shape of the axes over which normalization occurs, (W), (H,W) or (C,H,W)

NormalizedShape returns the collection as a Go slice.

func (*LayerNormalizationLayer) VarianceEpsilon added in v0.17.0 

func (lnl *LayerNormalizationLayer) VarianceEpsilon() float32

VarianceEpsilon returns a value used for numerical stability

func (*LayerNormalizationLayer) WithIsDebuggingEnabled added in v0.17.0 

func (lnl *LayerNormalizationLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *LayerNormalizationLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*LayerNormalizationLayer) WithLabel added in v0.17.0 

func (lnl *LayerNormalizationLayer) WithLabel(label string) *LayerNormalizationLayer

WithLabel sets a string that helps identify this layer.

type LayerProvider added in v0.17.0 

type LayerProvider interface {
	objref.Object
	// contains filtered or unexported methods
}

LayerProvider is accepted wherever a MLCLayer (or one of its subclasses) is expected.

type LossDescriptor added in v0.17.0 

type LossDescriptor struct {
	objref.Handle
}

LossDescriptor is an idiomatic wrapper over the Objective-C class MLCLossDescriptor.

A configuration object you use to create a loss layer.

func DescriptorWithTypeReductionType added in v0.17.0 

func DescriptorWithTypeReductionType(lossType LossType, reductionType ReductionType) *LossDescriptor

DescriptorWithTypeReductionType creates a loss descriptor with the loss function and reduction type you specify.

func DescriptorWithTypeReductionTypeWeight added in v0.17.0 

func DescriptorWithTypeReductionTypeWeight(lossType LossType, reductionType ReductionType, weight float32) *LossDescriptor

DescriptorWithTypeReductionTypeWeight creates a loss descriptor with the loss function, reduction type, and weight you specify.

func DescriptorWithTypeReductionTypeWeightLabelSmoothingClassCount added in v0.17.0 

func DescriptorWithTypeReductionTypeWeightLabelSmoothingClassCount(lossType LossType, reductionType ReductionType, weight float32, labelSmoothing float32, classCount int) *LossDescriptor

DescriptorWithTypeReductionTypeWeightLabelSmoothingClassCount creates a loss descriptor with the loss function, reduction type, weight, label smoothing, and number of classes you specify.

func DescriptorWithTypeReductionTypeWeightLabelSmoothingClassCountEpsilonDelta added in v0.17.0 

func DescriptorWithTypeReductionTypeWeightLabelSmoothingClassCountEpsilonDelta(lossType LossType, reductionType ReductionType, weight float32, labelSmoothing float32, classCount int, epsilon float32, delta float32) *LossDescriptor

DescriptorWithTypeReductionTypeWeightLabelSmoothingClassCountEpsilonDelta creates a loss descriptor with the loss function, reduction type, weight, label smoothing, and number of classes, epsilon, and delta that you specify.

func LossDescriptorFromID added in v0.17.0 

func LossDescriptorFromID(id objc.ID) *LossDescriptor

LossDescriptorFromID adopts an existing Objective-C object as a LossDescriptor (nil for 0), retaining it and registering a release finalizer.

func NewLossDescriptor added in v0.17.0 

func NewLossDescriptor() *LossDescriptor

NewLossDescriptor creates a new LossDescriptor.

func (*LossDescriptor) ClassCount added in v0.17.0 

func (ld *LossDescriptor) ClassCount() int

ClassCount returns the number of classes parameter. The default value is 1. This parameter is valid only for the loss function MLCLossTypeSoftmaxCrossEntropy.

func (*LossDescriptor) Delta added in v0.17.0 

func (ld *LossDescriptor) Delta() float32

Delta returns the delta parameter. The default value is 1.0f. This parameter is valid only for the loss function MLCLossTypeHuber.

func (*LossDescriptor) Description added in v0.17.0 

func (ld *LossDescriptor) Description() string

Description returns the object's -description text.

func (*LossDescriptor) Epsilon added in v0.17.0 

func (ld *LossDescriptor) Epsilon() float32

Epsilon returns the epsilon parameter. The default value is 1e-7. This parameter is valid only for the loss function MLCLossTypeLog.

func (*LossDescriptor) IsEqual added in v0.17.0 

func (ld *LossDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*LossDescriptor) IsKind added in v0.17.0 

func (ld *LossDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*LossDescriptor) LabelSmoothing added in v0.17.0 

func (ld *LossDescriptor) LabelSmoothing() float32

LabelSmoothing returns the label smoothing parameter. The default value is 0.0. This parameter is valid only for the loss functions of the following type(s): MLCLossTypeSoftmaxCrossEntropy and MLCLossTypeSigmoidCrossEntropy.

func (*LossDescriptor) LossType added in v0.17.0 

func (ld *LossDescriptor) LossType() LossType

LossType specifies the loss function.

func (*LossDescriptor) ReductionType added in v0.17.0 

func (ld *LossDescriptor) ReductionType() ReductionType

ReductionType returns the reduction operation performed by the loss function.

func (*LossDescriptor) String added in v0.17.0 

func (ld *LossDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*LossDescriptor) Weight added in v0.17.0 

func (ld *LossDescriptor) Weight() float32

Weight returns the scale factor to apply to each element of a result. The default value is 1.0.

type LossLayer added in v0.17.0 

type LossLayer struct {
	Layer
}

LossLayer is an idiomatic wrapper over the Objective-C class MLCLossLayer.

LossLayer is an abstract base — you do not construct it directly. Construct one of YOLOLossLayer and pass it where a LossLayer is accepted.

A layer that estimates the inaccuracies of the model to reduce the loss on the next evaluation.

func CategoricalCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeight added in v0.17.0 

func CategoricalCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeight(reductionType ReductionType, labelSmoothing float32, classCount int, weight float32) *LossLayer

CategoricalCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeight creates a categorical cross entropy loss layer with the reduction type, label smoothing, number of classes, and weight you specify.

func CategoricalCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeights added in v0.17.0 

func CategoricalCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeights(reductionType ReductionType, labelSmoothing float32, classCount int, weights *Tensor) *LossLayer

CategoricalCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeights creates a categorical cross entropy loss layer with the reduction type, label smoothing, number of classes, and weights you specify.

func CosineDistanceLossWithReductionTypeWeight added in v0.17.0 

func CosineDistanceLossWithReductionTypeWeight(reductionType ReductionType, weight float32) *LossLayer

CosineDistanceLossWithReductionTypeWeight creates a cosine distance loss layer with the reduction type and weight you specify.

func CosineDistanceLossWithReductionTypeWeights added in v0.17.0 

func CosineDistanceLossWithReductionTypeWeights(reductionType ReductionType, weights *Tensor) *LossLayer

CosineDistanceLossWithReductionTypeWeights creates a cosine distance loss layer with the reduction type and weights you specify.

func HingeLossWithReductionTypeWeight added in v0.17.0 

func HingeLossWithReductionTypeWeight(reductionType ReductionType, weight float32) *LossLayer

HingeLossWithReductionTypeWeight creates a hinge loss layer with the reduction type and weight you specify.

func HingeLossWithReductionTypeWeights added in v0.17.0 

func HingeLossWithReductionTypeWeights(reductionType ReductionType, weights *Tensor) *LossLayer

HingeLossWithReductionTypeWeights creates a hinge loss layer with the reduction type and weights you specify.

func HuberLossWithReductionTypeDeltaWeight added in v0.17.0 

func HuberLossWithReductionTypeDeltaWeight(reductionType ReductionType, delta float32, weight float32) *LossLayer

HuberLossWithReductionTypeDeltaWeight creates a huber loss layer with the reduction type, delta, and weight you specify.

func HuberLossWithReductionTypeDeltaWeights added in v0.17.0 

func HuberLossWithReductionTypeDeltaWeights(reductionType ReductionType, delta float32, weights *Tensor) *LossLayer

HuberLossWithReductionTypeDeltaWeights creates a huber loss layer with the reduction type, delta, and weights you specify.

func LogLossWithReductionTypeEpsilonWeight added in v0.17.0 

func LogLossWithReductionTypeEpsilonWeight(reductionType ReductionType, epsilon float32, weight float32) *LossLayer

LogLossWithReductionTypeEpsilonWeight creates a log loss layer with the reduction type, epsilon, and weight you specify.

func LogLossWithReductionTypeEpsilonWeights added in v0.17.0 

func LogLossWithReductionTypeEpsilonWeights(reductionType ReductionType, epsilon float32, weights *Tensor) *LossLayer

LogLossWithReductionTypeEpsilonWeights creates a log loss layer with the reduction type, epsilon, and weights you specify.

func LossLayerFromID added in v0.17.0 

func LossLayerFromID(id objc.ID) *LossLayer

LossLayerFromID adopts an existing Objective-C object as a LossLayer (nil for 0), retaining it and registering a release finalizer.

func MLCLossLayerLayerWithDescriptor 

func MLCLossLayerLayerWithDescriptor(lossDescriptor *LossDescriptor) *LossLayer

MLCLossLayerLayerWithDescriptor creates a loss layer with the descriptor you specify.

func MLCLossLayerLayerWithDescriptorWeights 

func MLCLossLayerLayerWithDescriptorWeights(lossDescriptor *LossDescriptor, weights *Tensor) *LossLayer

MLCLossLayerLayerWithDescriptorWeights creates a loss layer with the descriptor and weights you specify.

func MeanAbsoluteErrorLossWithReductionTypeWeight added in v0.17.0 

func MeanAbsoluteErrorLossWithReductionTypeWeight(reductionType ReductionType, weight float32) *LossLayer

MeanAbsoluteErrorLossWithReductionTypeWeight creates a mean absolute loss layer with the reduction type and weight.

func MeanAbsoluteErrorLossWithReductionTypeWeights added in v0.17.0 

func MeanAbsoluteErrorLossWithReductionTypeWeights(reductionType ReductionType, weights *Tensor) *LossLayer

MeanAbsoluteErrorLossWithReductionTypeWeights creates a mean absolute loss layer with the reduction type and weights you specify.

func MeanSquaredErrorLossWithReductionTypeWeight added in v0.17.0 

func MeanSquaredErrorLossWithReductionTypeWeight(reductionType ReductionType, weight float32) *LossLayer

MeanSquaredErrorLossWithReductionTypeWeight creates a mean squared loss layer with the reduction type and weight you specify.

func MeanSquaredErrorLossWithReductionTypeWeights added in v0.17.0 

func MeanSquaredErrorLossWithReductionTypeWeights(reductionType ReductionType, weights *Tensor) *LossLayer

MeanSquaredErrorLossWithReductionTypeWeights creates a mean squared loss layer with the reduction type and weights you specify.

func SigmoidCrossEntropyLossWithReductionTypeLabelSmoothingWeight added in v0.17.0 

func SigmoidCrossEntropyLossWithReductionTypeLabelSmoothingWeight(reductionType ReductionType, labelSmoothing float32, weight float32) *LossLayer

SigmoidCrossEntropyLossWithReductionTypeLabelSmoothingWeight creates a sigmoid cross entropy loss layer with the reduction type, label smoothing, and weight you specify.

func SigmoidCrossEntropyLossWithReductionTypeLabelSmoothingWeights added in v0.17.0 

func SigmoidCrossEntropyLossWithReductionTypeLabelSmoothingWeights(reductionType ReductionType, labelSmoothing float32, weights *Tensor) *LossLayer

SigmoidCrossEntropyLossWithReductionTypeLabelSmoothingWeights creates a sigmoid cross entropy loss layer with the reduction type, label smoothing, and weights you specify.

func SoftmaxCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeight added in v0.17.0 

func SoftmaxCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeight(reductionType ReductionType, labelSmoothing float32, classCount int, weight float32) *LossLayer

SoftmaxCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeight creates a softmax cross entropy loss layer with the reduction type, label smoothing, number of classes, and weight you specify.

func SoftmaxCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeights added in v0.17.0 

func SoftmaxCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeights(reductionType ReductionType, labelSmoothing float32, classCount int, weights *Tensor) *LossLayer

SoftmaxCrossEntropyLossWithReductionTypeLabelSmoothingClassCountWeights creates a softmax cross entropy loss layer with the reduction type, label smoothing, number of classes, and weights you specify.

func (*LossLayer) Descriptor added in v0.17.0 

func (ll *LossLayer) Descriptor() *LossDescriptor

Descriptor returns the loss descriptor

func (*LossLayer) Weights added in v0.17.0 

func (ll *LossLayer) Weights() *Tensor

Weights returns the loss label weights tensor

func (*LossLayer) WithIsDebuggingEnabled added in v0.17.0 

func (ll *LossLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *LossLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*LossLayer) WithLabel added in v0.17.0 

func (ll *LossLayer) WithLabel(label string) *LossLayer

WithLabel sets a string that helps identify this layer.

type LossLayerProvider added in v0.17.0 

type LossLayerProvider interface {
	objref.Object
	// contains filtered or unexported methods
}

LossLayerProvider is accepted wherever a MLCLossLayer (or one of its subclasses) is expected.

type LossType added in v0.17.0 

type LossType int32

A loss function. 

const (
	// The mean absolute error loss.
	LossTypeMeanAbsoluteError LossType = 0
	// The mean squared error loss.
	LossTypeMeanSquaredError LossType = 1
	// The softmax cross entropy loss.
	LossTypeSoftmaxCrossEntropy LossType = 2
	// The sigmoid cross entropy loss.
	LossTypeSigmoidCrossEntropy LossType = 3
	// The categorical cross entropy loss.
	LossTypeCategoricalCrossEntropy LossType = 4
	// The hinge loss.
	LossTypeHinge LossType = 5
	// The Huber loss.
	LossTypeHuber LossType = 6
	// The cosine distance loss.
	LossTypeCosineDistance LossType = 7
	// The log loss.
	LossTypeLog   LossType = 8
	LossTypeCount LossType = 9
)

func (LossType) String added in v0.17.0 

func (e LossType) String() string

String returns the LossType constant's name, or its numeric form when the value is not a known constant.

type MDLabelDomain 

type MDLabelDomain int32

These constants are used to specify a domain to MDLabelCreate(). 

const (
	KMDLabelUserDomain  MDLabelDomain = 0
	KMDLabelLocalDomain MDLabelDomain = 1
)

func (MDLabelDomain) String 

func (e MDLabelDomain) String() string

String returns the MDLabelDomain constant's name, or its numeric form when the value is not a known constant.

type MDQueryOptionFlags 

type MDQueryOptionFlags int32
const (
	KMDQuerySynchronous        MDQueryOptionFlags = 1
	KMDQueryWantsUpdates       MDQueryOptionFlags = 4
	KMDQueryAllowFSTranslation MDQueryOptionFlags = 8
)

func (MDQueryOptionFlags) String 

func (e MDQueryOptionFlags) String() string

String returns the MDQueryOptionFlags constant's name, or its numeric form when the value is not a known constant.

type MDQuerySortOptionFlags 

type MDQuerySortOptionFlags int32
const (
	KMDQueryReverseSortOrderFlag MDQuerySortOptionFlags = 1
)

func (MDQuerySortOptionFlags) String 

func (e MDQuerySortOptionFlags) String() string

String returns the MDQuerySortOptionFlags constant's name, or its numeric form when the value is not a known constant.

type MachVMRangeFlags added in v0.17.0 

type MachVMRangeFlags uint64

Bitmask — values may be combined with |. 

const (
	MachVMRangeFlagsNone MachVMRangeFlags = 0
)

func (MachVMRangeFlags) String added in v0.17.0 

func (e MachVMRangeFlags) String() string

String returns the MachVMRangeFlags constant's name, or its numeric form when the value is not a known constant.

type MachVMRangeFlavor added in v0.17.0 

type MachVMRangeFlavor uint32
const (
	MachVMRangeFlavorInvalid MachVMRangeFlavor = 0
	MachVMRangeFlavorV1      MachVMRangeFlavor = 1
)

func (MachVMRangeFlavor) String added in v0.17.0 

func (e MachVMRangeFlavor) String() string

String returns the MachVMRangeFlavor constant's name, or its numeric form when the value is not a known constant.

type MachVMRangeTag added in v0.17.0 

type MachVMRangeTag uint16
const (
	MachVMRangeTagDefault MachVMRangeTag = 0
	MachVMRangeTagData    MachVMRangeTag = 1
	MachVMRangeTagFixed   MachVMRangeTag = 2
)

func (MachVMRangeTag) String added in v0.17.0 

func (e MachVMRangeTag) String() string

String returns the MachVMRangeTag constant's name, or its numeric form when the value is not a known constant.

type MatMulDescriptor added in v0.17.0 

type MatMulDescriptor struct {
	objref.Handle
}

MatMulDescriptor is an idiomatic wrapper over the Objective-C class MLCMatMulDescriptor.

A configuration object you use to create a matrix multiplication layer.

func Descriptor added in v0.17.0 

func Descriptor() *MatMulDescriptor

Descriptor creates a batched matrix multiplication descriptor.

func DescriptorWithAlphaTransposesXTransposesY added in v0.17.0 

func DescriptorWithAlphaTransposesXTransposesY(alpha float32, transposesX bool, transposesY bool) *MatMulDescriptor

DescriptorWithAlphaTransposesXTransposesY creates a batched matrix multiplication descriptor with the alpha value and transpose options you specify.

func MatMulDescriptorFromID added in v0.17.0 

func MatMulDescriptorFromID(id objc.ID) *MatMulDescriptor

MatMulDescriptorFromID adopts an existing Objective-C object as a MatMulDescriptor (nil for 0), retaining it and registering a release finalizer.

func NewMatMulDescriptor added in v0.17.0 

func NewMatMulDescriptor() *MatMulDescriptor

NewMatMulDescriptor creates a new MatMulDescriptor.

func (*MatMulDescriptor) Alpha added in v0.17.0 

func (mmd *MatMulDescriptor) Alpha() float32

Alpha returns a scalar to scale the result in C=alpha x X x Y. Default = 1.0

func (*MatMulDescriptor) Description added in v0.17.0 

func (mmd *MatMulDescriptor) Description() string

Description returns the object's -description text.

func (*MatMulDescriptor) IsEqual added in v0.17.0 

func (mmd *MatMulDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*MatMulDescriptor) IsKind added in v0.17.0 

func (mmd *MatMulDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*MatMulDescriptor) String added in v0.17.0 

func (mmd *MatMulDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*MatMulDescriptor) TransposesX added in v0.17.0 

func (mmd *MatMulDescriptor) TransposesX() bool

TransposesX reports whether if true, transposes the last two dimensions of X. Default = False

func (*MatMulDescriptor) TransposesY added in v0.17.0 

func (mmd *MatMulDescriptor) TransposesY() bool

TransposesY reports whether if true, transposes the last two dimensions of Y. Default = False

type MatMulLayer added in v0.17.0 

type MatMulLayer struct {
	Layer
}

MatMulLayer is an idiomatic wrapper over the Objective-C class MLCMatMulLayer.

It embeds Layer, promoting that type's methods.

A layer that multiplies matrices.

func MLCMatMulLayerLayerWithDescriptor 

func MLCMatMulLayerLayerWithDescriptor(descriptor *MatMulDescriptor) *MatMulLayer

MLCMatMulLayerLayerWithDescriptor creates a matrix multiplication layer with the specified descriptor you specify.

func MatMulLayerFromID added in v0.17.0 

func MatMulLayerFromID(id objc.ID) *MatMulLayer

MatMulLayerFromID adopts an existing Objective-C object as a MatMulLayer (nil for 0), retaining it and registering a release finalizer.

func NewMatMulLayer added in v0.17.0 

func NewMatMulLayer() *MatMulLayer

NewMatMulLayer creates a new MatMulLayer.

func (*MatMulLayer) Descriptor added in v0.17.0 

func (mml *MatMulLayer) Descriptor() *MatMulDescriptor

Descriptor returns the matrix multiplication descriptor

func (*MatMulLayer) WithIsDebuggingEnabled added in v0.17.0 

func (mml *MatMulLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *MatMulLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*MatMulLayer) WithLabel added in v0.17.0 

func (mml *MatMulLayer) WithLabel(label string) *MatMulLayer

WithLabel sets a string that helps identify this layer.

type MpoFlags added in v0.17.0 

type MpoFlags uint32

Bitmask — values may be combined with |. 

const (
	MpoFlagsPort                        MpoFlags = 0
	MpoFlagsServicePort                 MpoFlags = 1024
	MpoFlagsConnectionPort              MpoFlags = 2048
	MpoFlagsReplyPort                   MpoFlags = 4096
	MpoFlagsWeakReplyPort               MpoFlags = 16384
	MpoFlagsNotificationPort            MpoFlags = 17408
	MpoFlagsExceptionPort               MpoFlags = 32768
	MpoFlagsConnectionPortWithPortArray MpoFlags = 65536
)

func (MpoFlags) String added in v0.17.0 

func (e MpoFlags) String() string

String returns the MpoFlags constant's name, or its numeric form when the value is not a known constant.

type MultiheadAttentionDescriptor added in v0.17.0 

type MultiheadAttentionDescriptor struct {
	objref.Handle
}

MultiheadAttentionDescriptor is an idiomatic wrapper over the Objective-C class MLCMultiheadAttentionDescriptor.

A configuration object you use to create a multi-head attention layer.

func DescriptorWithModelDimensionHeadCount added in v0.17.0 

func DescriptorWithModelDimensionHeadCount(modelDimension int, headCount int) *MultiheadAttentionDescriptor

DescriptorWithModelDimensionHeadCount creates a multi-head attention descriptor with the model dimension and number of parallel attention heads you specify.

func DescriptorWithModelDimensionKeyDimensionValueDimensionHeadCountDropoutHasBiasesHasAttentionBiasesAddsZeroAttention added in v0.17.0 

func DescriptorWithModelDimensionKeyDimensionValueDimensionHeadCountDropoutHasBiasesHasAttentionBiasesAddsZeroAttention(modelDimension int, keyDimension int, valueDimension int, headCount int, dropout float32, hasBiases bool, hasAttentionBiases bool, addsZeroAttention bool) *MultiheadAttentionDescriptor

DescriptorWithModelDimensionKeyDimensionValueDimensionHeadCountDropoutHasBiasesHasAttentionBiasesAddsZeroAttention creates a multi-head attention descriptor with the dimensions, number of attention heads, dropout rate, and bias and padding options you specify.

func MultiheadAttentionDescriptorFromID added in v0.17.0 

func MultiheadAttentionDescriptorFromID(id objc.ID) *MultiheadAttentionDescriptor

MultiheadAttentionDescriptorFromID adopts an existing Objective-C object as a MultiheadAttentionDescriptor (nil for 0), retaining it and registering a release finalizer.

func NewMultiheadAttentionDescriptor added in v0.17.0 

func NewMultiheadAttentionDescriptor() *MultiheadAttentionDescriptor

NewMultiheadAttentionDescriptor creates a new MultiheadAttentionDescriptor.

func (*MultiheadAttentionDescriptor) AddsZeroAttention added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) AddsZeroAttention() bool

AddsZeroAttention reports whether if true, a row of zeroes is added to projected key and value. Default = false

func (*MultiheadAttentionDescriptor) Description added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) Description() string

Description returns the object's -description text.

func (*MultiheadAttentionDescriptor) Dropout added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) Dropout() float32

Dropout returns a droupout layer applied to the output projection weights. Default = 0.0

func (*MultiheadAttentionDescriptor) HasAttentionBiases added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) HasAttentionBiases() bool

HasAttentionBiases reports whether if true, an array of biases is added to key and value respectively. Default = false

func (*MultiheadAttentionDescriptor) HasBiases added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) HasBiases() bool

HasBiases reports whether if true, bias is used for query/key/value/output projections. Default = true

func (*MultiheadAttentionDescriptor) HeadCount added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) HeadCount() int

HeadCount returns number of parallel attention heads

func (*MultiheadAttentionDescriptor) IsEqual added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*MultiheadAttentionDescriptor) IsKind added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*MultiheadAttentionDescriptor) KeyDimension added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) KeyDimension() int

KeyDimension returns total dimension of key space, Default = modelDimension

func (*MultiheadAttentionDescriptor) ModelDimension added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) ModelDimension() int

ModelDimension returns model or embedding dimension

func (*MultiheadAttentionDescriptor) String added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*MultiheadAttentionDescriptor) ValueDimension added in v0.17.0 

func (mad *MultiheadAttentionDescriptor) ValueDimension() int

ValueDimension returns total dimension of value space, Default = modelDimension

type MultiheadAttentionLayer added in v0.17.0 

type MultiheadAttentionLayer struct {
	Layer
}

MultiheadAttentionLayer is an idiomatic wrapper over the Objective-C class MLCMultiheadAttentionLayer.

It embeds Layer, promoting that type's methods.

A multihead, scaled dot-product attention layer that attends to one or more entries in the input key-value pairs.

func LayerWithDescriptorWeightsBiasesAttentionBiases added in v0.17.0 

func LayerWithDescriptorWeightsBiasesAttentionBiases(descriptor *MultiheadAttentionDescriptor, weights []*Tensor, biases []*Tensor, attentionBiases []*Tensor) *MultiheadAttentionLayer

LayerWithDescriptorWeightsBiasesAttentionBiases creates a multi-head attention layer with the descriptor, weights, and biases you specify.

func MultiheadAttentionLayerFromID added in v0.17.0 

func MultiheadAttentionLayerFromID(id objc.ID) *MultiheadAttentionLayer

MultiheadAttentionLayerFromID adopts an existing Objective-C object as a MultiheadAttentionLayer (nil for 0), retaining it and registering a release finalizer.

func NewMultiheadAttentionLayer added in v0.17.0 

func NewMultiheadAttentionLayer() *MultiheadAttentionLayer

NewMultiheadAttentionLayer creates a new MultiheadAttentionLayer.

func (*MultiheadAttentionLayer) AttentionBiases added in v0.17.0 

func (mal *MultiheadAttentionLayer) AttentionBiases() []*Tensor

AttentionBiases returns the biases added to key and value

AttentionBiases returns the collection as a Go slice.

func (*MultiheadAttentionLayer) Biases added in v0.17.0 

func (mal *MultiheadAttentionLayer) Biases() []*Tensor

Biases returns the biases of query, key, value and output projections

Biases returns the collection as a Go slice.

func (*MultiheadAttentionLayer) BiasesParameters added in v0.17.0 

func (mal *MultiheadAttentionLayer) BiasesParameters() []*TensorParameter

BiasesParameters returns the biases tensor parameters used for optimizer update

BiasesParameters returns the collection as a Go slice.

func (*MultiheadAttentionLayer) Descriptor added in v0.17.0 

func (mal *MultiheadAttentionLayer) Descriptor() *MultiheadAttentionDescriptor

Descriptor returns the multi-head attention descriptor

func (*MultiheadAttentionLayer) Weights added in v0.17.0 

func (mal *MultiheadAttentionLayer) Weights() []*Tensor

Weights returns the weights of query, key, value and output projections

Weights returns the collection as a Go slice.

func (*MultiheadAttentionLayer) WeightsParameters added in v0.17.0 

func (mal *MultiheadAttentionLayer) WeightsParameters() []*TensorParameter

WeightsParameters returns the weights tensor parameters used for optimizer update

WeightsParameters returns the collection as a Go slice.

func (*MultiheadAttentionLayer) WithIsDebuggingEnabled added in v0.17.0 

func (mal *MultiheadAttentionLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *MultiheadAttentionLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*MultiheadAttentionLayer) WithLabel added in v0.17.0 

func (mal *MultiheadAttentionLayer) WithLabel(label string) *MultiheadAttentionLayer

WithLabel sets a string that helps identify this layer.

type OSClockid added in v0.17.0 

type OSClockid uint32
const (
	OSClockidTime OSClockid = 32
)

func (OSClockid) String added in v0.17.0 

func (e OSClockid) String() string

String returns the OSClockid constant's name, or its numeric form when the value is not a known constant.

type Optimizer added in v0.17.0 

type Optimizer struct {
	objref.Handle
}

Optimizer is an idiomatic wrapper over the Objective-C class MLCOptimizer.

Optimizer is an abstract base — you do not construct it directly. Construct one of AdamOptimizer, AdamWOptimizer, RMSPropOptimizer, SGDOptimizer and pass it where a Optimizer is accepted.

The base class for all framework optimizers.

func OptimizerFromID added in v0.17.0 

func OptimizerFromID(id objc.ID) *Optimizer

OptimizerFromID adopts an existing Objective-C object as a Optimizer (nil for 0), retaining it and registering a release finalizer.

func (*Optimizer) AppliesGradientClipping added in v0.17.0 

func (o *Optimizer) AppliesGradientClipping() bool

AppliesGradientClipping reports whether gradient clipping should be applied or not.

func (*Optimizer) CustomGlobalNorm added in v0.17.0 

func (o *Optimizer) CustomGlobalNorm() float32

CustomGlobalNorm returns used only with MLCGradientClippingTypeByGlobalNorm. If non zero, this norm will be used in place of global norm.

func (*Optimizer) Description added in v0.17.0 

func (o *Optimizer) Description() string

Description returns the object's -description text.

func (*Optimizer) GradientClipMax added in v0.17.0 

func (o *Optimizer) GradientClipMax() float32

GradientClipMax returns the maximum gradient value if gradient clipping is enabled before gradient is rescaled.

func (*Optimizer) GradientClipMin added in v0.17.0 

func (o *Optimizer) GradientClipMin() float32

GradientClipMin returns the minimum gradient value if gradient clipping is enabled before gradient is rescaled.

func (*Optimizer) GradientClippingType added in v0.17.0 

func (o *Optimizer) GradientClippingType() GradientClippingType

GradientClippingType returns the type of clipping applied to gradient

func (*Optimizer) GradientRescale added in v0.17.0 

func (o *Optimizer) GradientRescale() float32

GradientRescale returns the rescale value applied to gradients during optimizer update

func (*Optimizer) IsEqual added in v0.17.0 

func (o *Optimizer) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*Optimizer) IsKind added in v0.17.0 

func (o *Optimizer) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*Optimizer) LearningRate added in v0.17.0 

func (o *Optimizer) LearningRate() float32

LearningRate returns the learning rate. This property is 'readwrite' so that callers can implement a 'decay' during training

func (*Optimizer) MaximumClippingNorm added in v0.17.0 

func (o *Optimizer) MaximumClippingNorm() float32

MaximumClippingNorm returns the maximum clipping value

func (*Optimizer) RegularizationScale added in v0.17.0 

func (o *Optimizer) RegularizationScale() float32

RegularizationScale returns the regularization scale.

func (*Optimizer) RegularizationType added in v0.17.0 

func (o *Optimizer) RegularizationType() RegularizationType

RegularizationType returns the regularization type.

func (*Optimizer) String added in v0.17.0 

func (o *Optimizer) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*Optimizer) WithAppliesGradientClipping added in v0.17.0 

func (o *Optimizer) WithAppliesGradientClipping(appliesGradientClipping bool) *Optimizer

WithAppliesGradientClipping sets a Boolean value that indicates whether you apply gradient clipping.

func (*Optimizer) WithLearningRate added in v0.17.0 

func (o *Optimizer) WithLearningRate(learningRate float32) *Optimizer

WithLearningRate sets the learning rate.

type OptimizerDescriptor added in v0.17.0 

type OptimizerDescriptor struct {
	objref.Handle
}

OptimizerDescriptor is an idiomatic wrapper over the Objective-C class MLCOptimizerDescriptor.

A configuration object you use to create an optimizer.

func DescriptorWithLearningRateGradientRescaleAppliesGradientClippingGradientClipMaxGradientClipMinRegularizationTypeRegularizationScale added in v0.17.0 

func DescriptorWithLearningRateGradientRescaleAppliesGradientClippingGradientClipMaxGradientClipMinRegularizationTypeRegularizationScale(learningRate float32, gradientRescale float32, appliesGradientClipping bool, gradientClipMax float32, gradientClipMin float32, regularizationType RegularizationType, regularizationScale float32) *OptimizerDescriptor

DescriptorWithLearningRateGradientRescaleAppliesGradientClippingGradientClipMaxGradientClipMinRegularizationTypeRegularizationScale creates a descriptor with the learning rate, gradient rescale, clipping option and values, and regularization type and scale that you specify.

func DescriptorWithLearningRateGradientRescaleAppliesGradientClippingGradientClippingTypeGradientClipMaxGradientClipMinMaximumClippingNormCustomGlobalNormRegularizationTypeRegularizationScale added in v0.17.0 

func DescriptorWithLearningRateGradientRescaleAppliesGradientClippingGradientClippingTypeGradientClipMaxGradientClipMinMaximumClippingNormCustomGlobalNormRegularizationTypeRegularizationScale(learningRate float32, gradientRescale float32, appliesGradientClipping bool, gradientClippingType GradientClippingType, gradientClipMax float32, gradientClipMin float32, maximumClippingNorm float32, customGlobalNorm float32, regularizationType RegularizationType, regularizationScale float32) *OptimizerDescriptor

DescriptorWithLearningRateGradientRescaleAppliesGradientClippingGradientClippingTypeGradientClipMaxGradientClipMinMaximumClippingNormCustomGlobalNormRegularizationTypeRegularizationScale creates a descriptor with the learning rate, gradient rescale, clipping option and values, and regularization type and scale that you specify.

func DescriptorWithLearningRateGradientRescaleRegularizationTypeRegularizationScale added in v0.17.0 

func DescriptorWithLearningRateGradientRescaleRegularizationTypeRegularizationScale(learningRate float32, gradientRescale float32, regularizationType RegularizationType, regularizationScale float32) *OptimizerDescriptor

DescriptorWithLearningRateGradientRescaleRegularizationTypeRegularizationScale creates an optimizer descriptor with the learning rate, gradient rescale, regularization type, and regulation scale that you specify.

func NewOptimizerDescriptor added in v0.17.0 

func NewOptimizerDescriptor() *OptimizerDescriptor

NewOptimizerDescriptor creates a new OptimizerDescriptor.

func OptimizerDescriptorFromID added in v0.17.0 

func OptimizerDescriptorFromID(id objc.ID) *OptimizerDescriptor

OptimizerDescriptorFromID adopts an existing Objective-C object as a OptimizerDescriptor (nil for 0), retaining it and registering a release finalizer.

func (*OptimizerDescriptor) AppliesGradientClipping added in v0.17.0 

func (od *OptimizerDescriptor) AppliesGradientClipping() bool

AppliesGradientClipping reports whether gradient clipping should be applied or not. The default is false

func (*OptimizerDescriptor) CustomGlobalNorm added in v0.17.0 

func (od *OptimizerDescriptor) CustomGlobalNorm() float32

CustomGlobalNorm returns used only with MLCGradientClippingTypeByGlobalNorm. If non zero, this norm will be used in place of global norm.

func (*OptimizerDescriptor) Description added in v0.17.0 

func (od *OptimizerDescriptor) Description() string

Description returns the object's -description text.

func (*OptimizerDescriptor) GradientClipMax added in v0.17.0 

func (od *OptimizerDescriptor) GradientClipMax() float32

GradientClipMax returns the maximum gradient value if gradient clipping is enabled before gradient is rescaled.

func (*OptimizerDescriptor) GradientClipMin added in v0.17.0 

func (od *OptimizerDescriptor) GradientClipMin() float32

GradientClipMin returns the minimum gradient value if gradient clipping is enabled before gradient is rescaled.

func (*OptimizerDescriptor) GradientClippingType added in v0.17.0 

func (od *OptimizerDescriptor) GradientClippingType() GradientClippingType

GradientClippingType returns the type of clipping applied to gradient

func (*OptimizerDescriptor) GradientRescale added in v0.17.0 

func (od *OptimizerDescriptor) GradientRescale() float32

GradientRescale returns the rescale value applied to gradients during optimizer update

func (*OptimizerDescriptor) IsEqual added in v0.17.0 

func (od *OptimizerDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*OptimizerDescriptor) IsKind added in v0.17.0 

func (od *OptimizerDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*OptimizerDescriptor) LearningRate added in v0.17.0 

func (od *OptimizerDescriptor) LearningRate() float32

LearningRate returns the learning rate

func (*OptimizerDescriptor) MaximumClippingNorm added in v0.17.0 

func (od *OptimizerDescriptor) MaximumClippingNorm() float32

MaximumClippingNorm returns the maximum clipping value

func (*OptimizerDescriptor) RegularizationScale added in v0.17.0 

func (od *OptimizerDescriptor) RegularizationScale() float32

RegularizationScale returns the regularization scale.

func (*OptimizerDescriptor) RegularizationType added in v0.17.0 

func (od *OptimizerDescriptor) RegularizationType() RegularizationType

RegularizationType returns the regularization type.

func (*OptimizerDescriptor) String added in v0.17.0 

func (od *OptimizerDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

type OptimizerProvider added in v0.17.0 

type OptimizerProvider interface {
	objref.Object
	// contains filtered or unexported methods
}

OptimizerProvider is accepted wherever a MLCOptimizer (or one of its subclasses) is expected.

type PaddingLayer added in v0.17.0 

type PaddingLayer struct {
	Layer
}

PaddingLayer is an idiomatic wrapper over the Objective-C class MLCPaddingLayer.

It embeds Layer, promoting that type's methods.

A layer that pads a tensor with the padding sizes you specify.

func LayerWithConstantPaddingConstantValue added in v0.17.0 

func LayerWithConstantPaddingConstantValue(padding []*foundation.Number, constantValue float32) *PaddingLayer

LayerWithConstantPaddingConstantValue creates a padding layer with the constant padding sizes and constant valu you specify.

func LayerWithReflectionPadding added in v0.17.0 

func LayerWithReflectionPadding(padding []*foundation.Number) *PaddingLayer

LayerWithReflectionPadding creates a padding layer with the reflection padding sizes you specify.

func LayerWithSymmetricPadding added in v0.17.0 

func LayerWithSymmetricPadding(padding []*foundation.Number) *PaddingLayer

LayerWithSymmetricPadding creates a padding layer with the symmetric padding sizes you specify.

func LayerWithZeroPadding added in v0.17.0 

func LayerWithZeroPadding(padding []*foundation.Number) *PaddingLayer

LayerWithZeroPadding creates a padding layer with the zero padding sizes you specify.

func NewPaddingLayer added in v0.17.0 

func NewPaddingLayer() *PaddingLayer

NewPaddingLayer creates a new PaddingLayer.

func PaddingLayerFromID added in v0.17.0 

func PaddingLayerFromID(id objc.ID) *PaddingLayer

PaddingLayerFromID adopts an existing Objective-C object as a PaddingLayer (nil for 0), retaining it and registering a release finalizer.

func (*PaddingLayer) ConstantValue added in v0.17.0 

func (pl *PaddingLayer) ConstantValue() float32

ConstantValue returns the constant value to use if padding type is constant.

func (*PaddingLayer) PaddingBottom added in v0.17.0 

func (pl *PaddingLayer) PaddingBottom() int

PaddingBottom returns the bottom padding size

func (*PaddingLayer) PaddingLeft added in v0.17.0 

func (pl *PaddingLayer) PaddingLeft() int

PaddingLeft returns the left padding size

func (*PaddingLayer) PaddingRight added in v0.17.0 

func (pl *PaddingLayer) PaddingRight() int

PaddingRight returns the right padding size

func (*PaddingLayer) PaddingTop added in v0.17.0 

func (pl *PaddingLayer) PaddingTop() int

PaddingTop returns the top padding size

func (*PaddingLayer) PaddingType added in v0.17.0 

func (pl *PaddingLayer) PaddingType() PaddingType

PaddingType returns the padding type i.e. constant, zero, reflect or symmetric

func (*PaddingLayer) WithIsDebuggingEnabled added in v0.17.0 

func (pl *PaddingLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *PaddingLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*PaddingLayer) WithLabel added in v0.17.0 

func (pl *PaddingLayer) WithLabel(label string) *PaddingLayer

WithLabel sets a string that helps identify this layer.

type PaddingPolicy added in v0.17.0 

type PaddingPolicy int32

A padding policy that you specify for a convolution or pooling layer. 

const (
	// The "same" padding policy.
	PaddingPolicySame PaddingPolicy = 0
	// The "valid" padding policy.
	PaddingPolicyValid PaddingPolicy = 1
	// The choice to use explicitly specified padding sizes.
	PaddingPolicyUsePaddingSize PaddingPolicy = 2
)

func (PaddingPolicy) String added in v0.17.0 

func (e PaddingPolicy) String() string

String returns the PaddingPolicy constant's name, or its numeric form when the value is not a known constant.

type PaddingType added in v0.17.0 

type PaddingType int32

A padding type that you specify for a padding layer. 

const (
	// The zero padding type.
	PaddingTypeZero PaddingType = 0
	// The reflect padding type.
	PaddingTypeReflect PaddingType = 1
	// The symmetric padding type.
	PaddingTypeSymmetric PaddingType = 2
	// The constant padding type.
	PaddingTypeConstant PaddingType = 3
)

func (PaddingType) String added in v0.17.0 

func (e PaddingType) String() string

String returns the PaddingType constant's name, or its numeric form when the value is not a known constant.

type Perm added in v0.17.0 

type Perm int32
const (
	PermReadData           Perm = 2
	PermListDirectory      Perm = 2
	PermWriteData          Perm = 4
	PermAddFile            Perm = 4
	PermExecute            Perm = 8
	PermSearch             Perm = 8
	PermDelete             Perm = 16
	PermAppendData         Perm = 32
	PermAddSubdirectory    Perm = 32
	PermDeleteChild        Perm = 64
	PermReadAttributes     Perm = 128
	PermWriteAttributes    Perm = 256
	PermReadExtattributes  Perm = 512
	PermWriteExtattributes Perm = 1024
	PermReadSecurity       Perm = 2048
	PermWriteSecurity      Perm = 4096
	PermChangeOwner        Perm = 8192
	PermSynchronize        Perm = 1048576
)

func (Perm) String added in v0.17.0 

func (e Perm) String() string

String returns the Perm constant's name, or its numeric form when the value is not a known constant.

type Platform added in v0.17.0 

type Platform struct {
	objref.Handle
}

Platform is an idiomatic wrapper over the Objective-C class MLCPlatform.

A utility class for setting global properties in the framework.

func NewPlatform added in v0.17.0 

func NewPlatform() *Platform

NewPlatform creates a new Platform.

func PlatformFromID added in v0.17.0 

func PlatformFromID(id objc.ID) *Platform

PlatformFromID adopts an existing Objective-C object as a Platform (nil for 0), retaining it and registering a release finalizer.

func (*Platform) Description added in v0.17.0 

func (p *Platform) Description() string

Description returns the object's -description text.

func (*Platform) IsEqual added in v0.17.0 

func (p *Platform) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*Platform) IsKind added in v0.17.0 

func (p *Platform) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*Platform) String added in v0.17.0 

func (p *Platform) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

type PoolingDescriptor added in v0.17.0 

type PoolingDescriptor struct {
	objref.Handle
}

PoolingDescriptor is an idiomatic wrapper over the Objective-C class MLCPoolingDescriptor.

A configuration object you use to create a pooling layer.

func AveragePoolingDescriptorWithKernelSizesStridesDilationRatesPaddingPolicyPaddingSizesCountIncludesPadding added in v0.17.0 

func AveragePoolingDescriptorWithKernelSizesStridesDilationRatesPaddingPolicyPaddingSizesCountIncludesPadding(kernelSizes []*foundation.Number, strides []*foundation.Number, dilationRates []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number, countIncludesPadding bool) *PoolingDescriptor

AveragePoolingDescriptorWithKernelSizesStridesDilationRatesPaddingPolicyPaddingSizesCountIncludesPadding creates an average pooling descriptor with the kernel sizes, strides, dilution rates, padding policy and sizes, and zero padding option you specify.

func AveragePoolingDescriptorWithKernelSizesStridesPaddingPolicyPaddingSizesCountIncludesPadding added in v0.17.0 

func AveragePoolingDescriptorWithKernelSizesStridesPaddingPolicyPaddingSizesCountIncludesPadding(kernelSizes []*foundation.Number, strides []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number, countIncludesPadding bool) *PoolingDescriptor

AveragePoolingDescriptorWithKernelSizesStridesPaddingPolicyPaddingSizesCountIncludesPadding creates an average pooling descriptor with the kernel sizes, strides, padding policy, padding sizes, and zero padding option that you specify.

func L2NormPoolingDescriptorWithKernelSizesStridesDilationRatesPaddingPolicyPaddingSizes added in v0.17.0 

func L2NormPoolingDescriptorWithKernelSizesStridesDilationRatesPaddingPolicyPaddingSizes(kernelSizes []*foundation.Number, strides []*foundation.Number, dilationRates []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *PoolingDescriptor

L2NormPoolingDescriptorWithKernelSizesStridesDilationRatesPaddingPolicyPaddingSizes creates a descriptor for an L2 norm pooling function with the kernel sizes, strides, dilation rates, padding policy, and padding sizes you specify.

func L2NormPoolingDescriptorWithKernelSizesStridesPaddingPolicyPaddingSizes added in v0.17.0 

func L2NormPoolingDescriptorWithKernelSizesStridesPaddingPolicyPaddingSizes(kernelSizes []*foundation.Number, strides []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *PoolingDescriptor

L2NormPoolingDescriptorWithKernelSizesStridesPaddingPolicyPaddingSizes creates a descriptor for an L2 norm pooling function with the kernel sizes, strides, padding policy, and padding sizes that you specify.

func MaxPoolingDescriptorWithKernelSizesStridesDilationRatesPaddingPolicyPaddingSizes added in v0.17.0 

func MaxPoolingDescriptorWithKernelSizesStridesDilationRatesPaddingPolicyPaddingSizes(kernelSizes []*foundation.Number, strides []*foundation.Number, dilationRates []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *PoolingDescriptor

MaxPoolingDescriptorWithKernelSizesStridesDilationRatesPaddingPolicyPaddingSizes creates a descriptor for a max pooling function with the kernel sizes, strides, dilation rates, padding policy, and padding sizes that you specify.

func MaxPoolingDescriptorWithKernelSizesStridesPaddingPolicyPaddingSizes added in v0.17.0 

func MaxPoolingDescriptorWithKernelSizesStridesPaddingPolicyPaddingSizes(kernelSizes []*foundation.Number, strides []*foundation.Number, paddingPolicy PaddingPolicy, paddingSizes []*foundation.Number) *PoolingDescriptor

MaxPoolingDescriptorWithKernelSizesStridesPaddingPolicyPaddingSizes creates a descriptor for a max pooling function with the kernel sizes, strides, padding policy, and padding sizes that you specify.

func NewPoolingDescriptor added in v0.17.0 

func NewPoolingDescriptor() *PoolingDescriptor

NewPoolingDescriptor creates a new PoolingDescriptor.

func PoolingDescriptorFromID added in v0.17.0 

func PoolingDescriptorFromID(id objc.ID) *PoolingDescriptor

PoolingDescriptorFromID adopts an existing Objective-C object as a PoolingDescriptor (nil for 0), retaining it and registering a release finalizer.

func PoolingDescriptorWithTypeKernelSizeStride added in v0.17.0 

func PoolingDescriptorWithTypeKernelSizeStride(poolingType PoolingType, kernelSize int, stride int) *PoolingDescriptor

PoolingDescriptorWithTypeKernelSizeStride creates a pooling descriptor with the pooling function, kernel size, and stride you specify.

func (*PoolingDescriptor) CountIncludesPadding added in v0.17.0 

func (pd *PoolingDescriptor) CountIncludesPadding() bool

CountIncludesPadding reports whether include the zero-padding in the averaging calculation if true. Used only with average pooling.

func (*PoolingDescriptor) Description added in v0.17.0 

func (pd *PoolingDescriptor) Description() string

Description returns the object's -description text.

func (*PoolingDescriptor) DilationRateInX added in v0.17.0 

func (pd *PoolingDescriptor) DilationRateInX() int

DilationRateInX returns the dilation rate i.e. stride of elements in the kernel in x.

func (*PoolingDescriptor) DilationRateInY added in v0.17.0 

func (pd *PoolingDescriptor) DilationRateInY() int

DilationRateInY returns the dilation rate i.e. stride of elements in the kernel in y.

func (*PoolingDescriptor) IsEqual added in v0.17.0 

func (pd *PoolingDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*PoolingDescriptor) IsKind added in v0.17.0 

func (pd *PoolingDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*PoolingDescriptor) KernelHeight added in v0.17.0 

func (pd *PoolingDescriptor) KernelHeight() int

KernelHeight returns the pooling kernel size in y.

func (*PoolingDescriptor) KernelWidth added in v0.17.0 

func (pd *PoolingDescriptor) KernelWidth() int

KernelWidth returns the pooling kernel size in x.

func (*PoolingDescriptor) PaddingPolicy added in v0.17.0 

func (pd *PoolingDescriptor) PaddingPolicy() PaddingPolicy

PaddingPolicy returns the padding policy to use.

func (*PoolingDescriptor) PaddingSizeInX added in v0.17.0 

func (pd *PoolingDescriptor) PaddingSizeInX() int

PaddingSizeInX returns the padding size in x (left and right) to use if paddingPolicy is MLCPaddingPolicyUsePaddingSize

func (*PoolingDescriptor) PaddingSizeInY added in v0.17.0 

func (pd *PoolingDescriptor) PaddingSizeInY() int

PaddingSizeInY returns the padding size in y (top and bottom) to use if paddingPolicy is MLCPaddingPolicyUsePaddingSize

func (*PoolingDescriptor) PoolingType added in v0.17.0 

func (pd *PoolingDescriptor) PoolingType() PoolingType

PoolingType returns the pooling operation

func (*PoolingDescriptor) StrideInX added in v0.17.0 

func (pd *PoolingDescriptor) StrideInX() int

StrideInX returns the stride of the kernel in x.

func (*PoolingDescriptor) StrideInY added in v0.17.0 

func (pd *PoolingDescriptor) StrideInY() int

StrideInY returns the stride of the kernel in y.

func (*PoolingDescriptor) String added in v0.17.0 

func (pd *PoolingDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

type PoolingLayer added in v0.17.0 

type PoolingLayer struct {
	Layer
}

PoolingLayer is an idiomatic wrapper over the Objective-C class MLCPoolingLayer.

It embeds Layer, promoting that type's methods.

A layer that summarizes the average presence of a feature.

func MLCPoolingLayerLayerWithDescriptor 

func MLCPoolingLayerLayerWithDescriptor(descriptor *PoolingDescriptor) *PoolingLayer

MLCPoolingLayerLayerWithDescriptor creates a pooling layer with the descriptor you specify.

func NewPoolingLayer added in v0.17.0 

func NewPoolingLayer() *PoolingLayer

NewPoolingLayer creates a new PoolingLayer.

func PoolingLayerFromID added in v0.17.0 

func PoolingLayerFromID(id objc.ID) *PoolingLayer

PoolingLayerFromID adopts an existing Objective-C object as a PoolingLayer (nil for 0), retaining it and registering a release finalizer.

func (*PoolingLayer) Descriptor added in v0.17.0 

func (pl *PoolingLayer) Descriptor() *PoolingDescriptor

Descriptor returns the pooling descriptor

func (*PoolingLayer) WithIsDebuggingEnabled added in v0.17.0 

func (pl *PoolingLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *PoolingLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*PoolingLayer) WithLabel added in v0.17.0 

func (pl *PoolingLayer) WithLabel(label string) *PoolingLayer

WithLabel sets a string that helps identify this layer.

type PoolingType added in v0.17.0 

type PoolingType int32

A pooling function type for a pooling layer. 

const (
	// The max pooling type.
	PoolingTypeMax PoolingType = 1
	// The average pooling type.
	PoolingTypeAverage PoolingType = 2
	// The L2-norm pooling type.
	PoolingTypeL2Norm PoolingType = 3
	PoolingTypeCount  PoolingType = 4
)

func (PoolingType) String added in v0.17.0 

func (e PoolingType) String() string

String returns the PoolingType constant's name, or its numeric form when the value is not a known constant.

type PtrauthKey added in v0.17.0 

type PtrauthKey int32
const (
	Ptrauth_key_none                     PtrauthKey = -1
	Ptrauth_key_asia                     PtrauthKey = 0
	Ptrauth_key_asib                     PtrauthKey = 1
	Ptrauth_key_asda                     PtrauthKey = 2
	Ptrauth_key_asdb                     PtrauthKey = 3
	Ptrauth_key_process_independent_code PtrauthKey = 0
	Ptrauth_key_process_dependent_code   PtrauthKey = 1
	Ptrauth_key_process_independent_data PtrauthKey = 2
	Ptrauth_key_process_dependent_data   PtrauthKey = 3
	Ptrauth_key_return_address           PtrauthKey = 1
	Ptrauth_key_frame_pointer            PtrauthKey = 3
	Ptrauth_key_function_pointer         PtrauthKey = 0
	Ptrauth_key_block_function           PtrauthKey = 0
	Ptrauth_key_cxx_vtable_pointer       PtrauthKey = 2
	Ptrauth_key_method_list_pointer      PtrauthKey = 2
	Ptrauth_key_objc_isa_pointer         PtrauthKey = 2
	Ptrauth_key_objc_super_pointer       PtrauthKey = 2
	Ptrauth_key_objc_sel_pointer         PtrauthKey = 3
	Ptrauth_key_objc_class_ro_pointer    PtrauthKey = 2
	Ptrauth_key_block_descriptor_pointer PtrauthKey = 2
	Ptrauth_key_init_fini_pointer        PtrauthKey = 0
)

func (PtrauthKey) String added in v0.17.0 

func (e PtrauthKey) String() string

String returns the PtrauthKey constant's name, or its numeric form when the value is not a known constant.

type QosClass added in v0.17.0 

type QosClass uint32
const (
	QosClassUserInteractive QosClass = 33
	QosClassUserInitiated   QosClass = 25
	QosClassDefault         QosClass = 21
	QosClassUtility         QosClass = 17
	QosClassBackground      QosClass = 9
	QosClassUnspecified     QosClass = 0
)

func (QosClass) String added in v0.17.0 

func (e QosClass) String() string

String returns the QosClass constant's name, or its numeric form when the value is not a known constant.

type RMSPropOptimizer added in v0.17.0 

type RMSPropOptimizer struct {
	Optimizer
}

RMSPropOptimizer is an idiomatic wrapper over the Objective-C class MLCRMSPropOptimizer.

It embeds Optimizer, promoting that type's methods.

An optimizer that represents the root mean square propagation algorithm.

func MLCRMSPropOptimizerOptimizerWithDescriptor 

func MLCRMSPropOptimizerOptimizerWithDescriptor(optimizerDescriptor *OptimizerDescriptor) *RMSPropOptimizer

MLCRMSPropOptimizerOptimizerWithDescriptor creates an RMSProp optimizer with the descriptor you specify.

func NewRMSPropOptimizer added in v0.17.0 

func NewRMSPropOptimizer() *RMSPropOptimizer

NewRMSPropOptimizer creates a new RMSPropOptimizer.

func OptimizerWithDescriptorMomentumScaleAlphaEpsilonIsCentered added in v0.17.0 

func OptimizerWithDescriptorMomentumScaleAlphaEpsilonIsCentered(optimizerDescriptor *OptimizerDescriptor, momentumScale float32, alpha float32, epsilon float32, isCentered bool) *RMSPropOptimizer

OptimizerWithDescriptorMomentumScaleAlphaEpsilonIsCentered creates an RMSProp optimizer with the descriptor, momentum scale, smoothing, epsilon, and option to compute the centered RMSProp that you specify.

func RMSPropOptimizerFromID added in v0.17.0 

func RMSPropOptimizerFromID(id objc.ID) *RMSPropOptimizer

RMSPropOptimizerFromID adopts an existing Objective-C object as a RMSPropOptimizer (nil for 0), retaining it and registering a release finalizer.

func (*RMSPropOptimizer) Alpha added in v0.17.0 

func (rpo *RMSPropOptimizer) Alpha() float32

Alpha returns the smoothing constant. The default is 0.99.

func (*RMSPropOptimizer) Epsilon added in v0.17.0 

func (rpo *RMSPropOptimizer) Epsilon() float32

Epsilon returns a term added to improve numerical stability. The default is 1e-8.

func (*RMSPropOptimizer) IsCentered added in v0.17.0 

func (rpo *RMSPropOptimizer) IsCentered() bool

IsCentered reports whether if True, compute the centered RMSProp, the gradient is normalized by an estimation of its variance. The default is false.

func (*RMSPropOptimizer) MomentumScale added in v0.17.0 

func (rpo *RMSPropOptimizer) MomentumScale() float32

MomentumScale returns the momentum factor. A hyper-parameter. The default is 0.0.

func (*RMSPropOptimizer) WithAppliesGradientClipping added in v0.17.0 

func (rpo *RMSPropOptimizer) WithAppliesGradientClipping(appliesGradientClipping bool) *RMSPropOptimizer

WithAppliesGradientClipping sets a Boolean value that indicates whether you apply gradient clipping.

func (*RMSPropOptimizer) WithLearningRate added in v0.17.0 

func (rpo *RMSPropOptimizer) WithLearningRate(learningRate float32) *RMSPropOptimizer

WithLearningRate sets the learning rate.

type RandomInitializerType added in v0.17.0 

type RandomInitializerType int32

An initializer type you use to create a tensor with random data. 

const (
	RandomInitializerTypeInvalid RandomInitializerType = 0
	// The uniform random initializer type.
	RandomInitializerTypeUniform RandomInitializerType = 1
	// The glorot uniform random initializer type.
	RandomInitializerTypeGlorotUniform RandomInitializerType = 2
	// The Xavier random initializer type.
	RandomInitializerTypeXavier RandomInitializerType = 3
	RandomInitializerTypeCount  RandomInitializerType = 4
)

func (RandomInitializerType) String added in v0.17.0 

func (e RandomInitializerType) String() string

String returns the RandomInitializerType constant's name, or its numeric form when the value is not a known constant.

type ReductionLayer added in v0.17.0 

type ReductionLayer struct {
	Layer
}

ReductionLayer is an idiomatic wrapper over the Objective-C class MLCReductionLayer.

It embeds Layer, promoting that type's methods.

A layer that reduces tensor values across a specific dimension to a scalar value.

func LayerWithReductionTypeDimension added in v0.17.0 

func LayerWithReductionTypeDimension(reductionType ReductionType, dimension int) *ReductionLayer

LayerWithReductionTypeDimension creates a reduction layer using the reduction type and dimension you specify.

func LayerWithReductionTypeDimensions added in v0.17.0 

func LayerWithReductionTypeDimensions(reductionType ReductionType, dimensions []*foundation.Number) *ReductionLayer

LayerWithReductionTypeDimensions creates a reduction layer using the reduction type and dimensions you specify.

func NewReductionLayer added in v0.17.0 

func NewReductionLayer() *ReductionLayer

NewReductionLayer creates a new ReductionLayer.

func ReductionLayerFromID added in v0.17.0 

func ReductionLayerFromID(id objc.ID) *ReductionLayer

ReductionLayerFromID adopts an existing Objective-C object as a ReductionLayer (nil for 0), retaining it and registering a release finalizer.

func (*ReductionLayer) Dimension added in v0.17.0 

func (rl *ReductionLayer) Dimension() int

Dimension returns the dimension over which to perform the reduction operation

func (*ReductionLayer) Dimensions added in v0.17.0 

func (rl *ReductionLayer) Dimensions() []obj.Object

Dimensions returns the dimensions over which to perform the reduction operation

Dimensions returns the collection as a Go slice.

func (*ReductionLayer) ReductionType added in v0.17.0 

func (rl *ReductionLayer) ReductionType() ReductionType

ReductionType returns the reduction type

func (*ReductionLayer) WithIsDebuggingEnabled added in v0.17.0 

func (rl *ReductionLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *ReductionLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*ReductionLayer) WithLabel added in v0.17.0 

func (rl *ReductionLayer) WithLabel(label string) *ReductionLayer

WithLabel sets a string that helps identify this layer.

type ReductionType added in v0.17.0 

type ReductionType int32

Constants that describe a reduction operation type. 

const (
	// A reduction operation that applies no reduction.
	ReductionTypeNone ReductionType = 0
	// A reduction operation that applies to the sum of the dimensions.
	ReductionTypeSum ReductionType = 1
	// A reduction operation that applies to the mean of the dimensions.
	ReductionTypeMean ReductionType = 2
	// A reduction operation that applies to the maximum dimension.
	ReductionTypeMax ReductionType = 3
	// A reduction operation that applies to the minimum dimension.
	ReductionTypeMin ReductionType = 4
	// A reduction operation that applies to the maximum dimension you specify.
	ReductionTypeArgMax ReductionType = 5
	// A reduction operation that applies to the minimum dimension you specify.
	ReductionTypeArgMin ReductionType = 6
	// A reduction operation that applies a lasso regularization penalty.
	ReductionTypeL1Norm ReductionType = 7
	// A reduction operation that applies to any dimension.
	ReductionTypeAny ReductionType = 8
	// A reduction operation that applies to all dimensions.
	ReductionTypeAll ReductionType = 9
	// The total number of reduction operations.
	ReductionTypeCount ReductionType = 10
)

func (ReductionType) String added in v0.17.0 

func (e ReductionType) String() string

String returns the ReductionType constant's name, or its numeric form when the value is not a known constant.

type RegularizationType added in v0.17.0 

type RegularizationType int32

A regularization function to use with an optimizer. 

const (
	// No regularization.
	RegularizationTypeNone RegularizationType = 0
	// The L1 regularization.
	RegularizationTypeL1 RegularizationType = 1
	// The L2 regularization.
	RegularizationTypeL2 RegularizationType = 2
)

func (RegularizationType) String added in v0.17.0 

func (e RegularizationType) String() string

String returns the RegularizationType constant's name, or its numeric form when the value is not a known constant.

type ReshapeLayer added in v0.17.0 

type ReshapeLayer struct {
	Layer
}

ReshapeLayer is an idiomatic wrapper over the Objective-C class MLCReshapeLayer.

It embeds Layer, promoting that type's methods.

A layer that reshapes a tensor with the shape you specify.

func LayerWithShape added in v0.17.0 

func LayerWithShape(shape []*foundation.Number) *ReshapeLayer

LayerWithShape creates a reshape layer with the shape you specify.

func NewReshapeLayer added in v0.17.0 

func NewReshapeLayer() *ReshapeLayer

NewReshapeLayer creates a new ReshapeLayer.

func ReshapeLayerFromID added in v0.17.0 

func ReshapeLayerFromID(id objc.ID) *ReshapeLayer

ReshapeLayerFromID adopts an existing Objective-C object as a ReshapeLayer (nil for 0), retaining it and registering a release finalizer.

func (*ReshapeLayer) Shape added in v0.17.0 

func (rl *ReshapeLayer) Shape() []obj.Object

Shape returns the target shape.

Shape returns the collection as a Go slice.

func (*ReshapeLayer) WithIsDebuggingEnabled added in v0.17.0 

func (rl *ReshapeLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *ReshapeLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*ReshapeLayer) WithLabel added in v0.17.0 

func (rl *ReshapeLayer) WithLabel(label string) *ReshapeLayer

WithLabel sets a string that helps identify this layer.

type SGDOptimizer added in v0.17.0 

type SGDOptimizer struct {
	Optimizer
}

SGDOptimizer is an idiomatic wrapper over the Objective-C class MLCSGDOptimizer.

It embeds Optimizer, promoting that type's methods.

An optimizer that represents the stochastic gradient decent algorithm.

func MLCSGDOptimizerOptimizerWithDescriptor 

func MLCSGDOptimizerOptimizerWithDescriptor(optimizerDescriptor *OptimizerDescriptor) *SGDOptimizer

MLCSGDOptimizerOptimizerWithDescriptor creates an SGD optimizer with the descriptor you specify.

func NewSGDOptimizer added in v0.17.0 

func NewSGDOptimizer() *SGDOptimizer

NewSGDOptimizer creates a new SGDOptimizer.

func OptimizerWithDescriptorMomentumScaleUsesNesterovMomentum added in v0.17.0 

func OptimizerWithDescriptorMomentumScaleUsesNesterovMomentum(optimizerDescriptor *OptimizerDescriptor, momentumScale float32, usesNesterovMomentum bool) *SGDOptimizer

OptimizerWithDescriptorMomentumScaleUsesNesterovMomentum create an SGD optimizer with the descriptor, momentum scale, and option to enable Nesterov momentum that you specify.

func SGDOptimizerFromID added in v0.17.0 

func SGDOptimizerFromID(id objc.ID) *SGDOptimizer

SGDOptimizerFromID adopts an existing Objective-C object as a SGDOptimizer (nil for 0), retaining it and registering a release finalizer.

func (*SGDOptimizer) MomentumScale added in v0.17.0 

func (so *SGDOptimizer) MomentumScale() float32

MomentumScale returns the momentum factor. A hyper-parameter. The default is 0.0.

func (*SGDOptimizer) UsesNesterovMomentum added in v0.17.0 

func (so *SGDOptimizer) UsesNesterovMomentum() bool

UsesNesterovMomentum reports whether a boolean that specifies whether to apply nesterov momentum or not. The default is false.

func (*SGDOptimizer) WithAppliesGradientClipping added in v0.17.0 

func (so *SGDOptimizer) WithAppliesGradientClipping(appliesGradientClipping bool) *SGDOptimizer

WithAppliesGradientClipping sets a Boolean value that indicates whether you apply gradient clipping.

func (*SGDOptimizer) WithLearningRate added in v0.17.0 

func (so *SGDOptimizer) WithLearningRate(learningRate float32) *SGDOptimizer

WithLearningRate sets the learning rate.

type SampleMode added in v0.17.0 

type SampleMode int32

A sampling mode for an upsample layer. 

const (
	// The nearest sample mode.
	SampleModeNearest SampleMode = 0
	// The linear sample mode.
	SampleModeLinear SampleMode = 1
)

func (SampleMode) String added in v0.17.0 

func (e SampleMode) String() string

String returns the SampleMode constant's name, or its numeric form when the value is not a known constant.

type ScatterLayer added in v0.17.0 

type ScatterLayer struct {
	Layer
}

ScatterLayer is an idiomatic wrapper over the Objective-C class MLCScatterLayer.

It embeds Layer, promoting that type's methods.

A layer that updates the output at an index you specify.

func LayerWithDimensionReductionType added in v0.17.0 

func LayerWithDimensionReductionType(dimension int, reductionType ReductionType) *ScatterLayer

LayerWithDimensionReductionType creates a scatter layer with the dimension and reduction type you specify.

func NewScatterLayer added in v0.17.0 

func NewScatterLayer() *ScatterLayer

NewScatterLayer creates a new ScatterLayer.

func ScatterLayerFromID added in v0.17.0 

func ScatterLayerFromID(id objc.ID) *ScatterLayer

ScatterLayerFromID adopts an existing Objective-C object as a ScatterLayer (nil for 0), retaining it and registering a release finalizer.

func (*ScatterLayer) Dimension added in v0.17.0 

func (sl *ScatterLayer) Dimension() int

Dimension returns the dimension along which to index

func (*ScatterLayer) ReductionType added in v0.17.0 

func (sl *ScatterLayer) ReductionType() ReductionType

ReductionType returns the reduction type applied for all values in source tensor that are scattered to a specific location in the result tensor. Must be: MLCReductionTypeNone or MLCReductionTypeSum.

func (*ScatterLayer) WithIsDebuggingEnabled added in v0.17.0 

func (sl *ScatterLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *ScatterLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*ScatterLayer) WithLabel added in v0.17.0 

func (sl *ScatterLayer) WithLabel(label string) *ScatterLayer

WithLabel sets a string that helps identify this layer.

type SelectionLayer added in v0.17.0 

type SelectionLayer struct {
	Layer
}

SelectionLayer is an idiomatic wrapper over the Objective-C class MLCSelectionLayer.

It embeds Layer, promoting that type's methods.

A layer for selecting elements from two tensors.

func MLCSelectionLayerLayer 

func MLCSelectionLayerLayer() *SelectionLayer

MLCSelectionLayerLayer creates a selection layer.

func NewSelectionLayer added in v0.17.0 

func NewSelectionLayer() *SelectionLayer

NewSelectionLayer creates a new SelectionLayer.

func SelectionLayerFromID added in v0.17.0 

func SelectionLayerFromID(id objc.ID) *SelectionLayer

SelectionLayerFromID adopts an existing Objective-C object as a SelectionLayer (nil for 0), retaining it and registering a release finalizer.

func (*SelectionLayer) WithIsDebuggingEnabled added in v0.17.0 

func (sl *SelectionLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *SelectionLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*SelectionLayer) WithLabel added in v0.17.0 

func (sl *SelectionLayer) WithLabel(label string) *SelectionLayer

WithLabel sets a string that helps identify this layer.

type SliceLayer added in v0.17.0 

type SliceLayer struct {
	Layer
}

SliceLayer is an idiomatic wrapper over the Objective-C class MLCSliceLayer.

It embeds Layer, promoting that type's methods.

A layer that extracts a slice from a tensor.

func NewSliceLayer added in v0.17.0 

func NewSliceLayer() *SliceLayer

NewSliceLayer creates a new SliceLayer.

func SliceLayerFromID added in v0.17.0 

func SliceLayerFromID(id objc.ID) *SliceLayer

SliceLayerFromID adopts an existing Objective-C object as a SliceLayer (nil for 0), retaining it and registering a release finalizer.

func SliceLayerWithStartEndStride added in v0.17.0 

func SliceLayerWithStartEndStride(start []*foundation.Number, end []*foundation.Number, stride []*foundation.Number) *SliceLayer

SliceLayerWithStartEndStride creates a slice layer with the specified start, end, and stride.

func (*SliceLayer) End added in v0.17.0 

func (sl *SliceLayer) End() []obj.Object

End returns a vector of length equal to that of source. The element at index i specifies the end of slice in dimension i.

End returns the collection as a Go slice.

func (*SliceLayer) Start added in v0.17.0 

func (sl *SliceLayer) Start() []obj.Object

Start returns a vector of length equal to that of source. The element at index i specifies the beginning of slice in dimension i.

Start returns the collection as a Go slice.

func (*SliceLayer) Stride added in v0.17.0 

func (sl *SliceLayer) Stride() []obj.Object

Stride returns a vector of length equal to that of source. The element at index i specifies the stride of slice in dimension i.

Stride returns the collection as a Go slice.

func (*SliceLayer) WithIsDebuggingEnabled added in v0.17.0 

func (sl *SliceLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *SliceLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*SliceLayer) WithLabel added in v0.17.0 

func (sl *SliceLayer) WithLabel(label string) *SliceLayer

WithLabel sets a string that helps identify this layer.

type SoftmaxLayer added in v0.17.0 

type SoftmaxLayer struct {
	Layer
}

SoftmaxLayer is an idiomatic wrapper over the Objective-C class MLCSoftmaxLayer.

It embeds Layer, promoting that type's methods.

A layer that outputs a probability distribution as attention weights.

func LayerWithOperationDimension added in v0.17.0 

func LayerWithOperationDimension(operation SoftmaxOperation, dimension int) *SoftmaxLayer

LayerWithOperationDimension creates a softmax layer with the operation and dimension you specify.

func MLCSoftmaxLayerLayerWithOperation 

func MLCSoftmaxLayerLayerWithOperation(operation SoftmaxOperation) *SoftmaxLayer

MLCSoftmaxLayerLayerWithOperation creates a softmax layer with the operation you specify.

func NewSoftmaxLayer added in v0.17.0 

func NewSoftmaxLayer() *SoftmaxLayer

NewSoftmaxLayer creates a new SoftmaxLayer.

func SoftmaxLayerFromID added in v0.17.0 

func SoftmaxLayerFromID(id objc.ID) *SoftmaxLayer

SoftmaxLayerFromID adopts an existing Objective-C object as a SoftmaxLayer (nil for 0), retaining it and registering a release finalizer.

func (*SoftmaxLayer) Dimension added in v0.17.0 

func (sl *SoftmaxLayer) Dimension() int

Dimension returns the dimension over which softmax operation should be performed

func (*SoftmaxLayer) Operation added in v0.17.0 

func (sl *SoftmaxLayer) Operation() SoftmaxOperation

Operation returns the softmax operation. Supported values are softmax and log softmax.

func (*SoftmaxLayer) WithIsDebuggingEnabled added in v0.17.0 

func (sl *SoftmaxLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *SoftmaxLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*SoftmaxLayer) WithLabel added in v0.17.0 

func (sl *SoftmaxLayer) WithLabel(label string) *SoftmaxLayer

WithLabel sets a string that helps identify this layer.

type SoftmaxOperation added in v0.17.0 

type SoftmaxOperation int32

A softmax operation. 

const (
	// The standard softmax operation.
	SoftmaxOperationSoftmax SoftmaxOperation = 0
	// The log softmax operation.
	SoftmaxOperationLogSoftmax SoftmaxOperation = 1
)

func (SoftmaxOperation) String added in v0.17.0 

func (e SoftmaxOperation) String() string

String returns the SoftmaxOperation constant's name, or its numeric form when the value is not a known constant.

type SplitLayer added in v0.17.0 

type SplitLayer struct {
	Layer
}

SplitLayer is an idiomatic wrapper over the Objective-C class MLCSplitLayer.

It embeds Layer, promoting that type's methods.

A layer that splits a tensor value into a list of subtensors.

func LayerWithSplitCountDimension added in v0.17.0 

func LayerWithSplitCountDimension(splitCount int, dimension int) *SplitLayer

LayerWithSplitCountDimension creates a split layer with the number of splits and dimension you specify.

func LayerWithSplitSectionLengthsDimension added in v0.17.0 

func LayerWithSplitSectionLengthsDimension(splitSectionLengths []*foundation.Number, dimension int) *SplitLayer

LayerWithSplitSectionLengthsDimension creates a split layer with the lengths of each split section and dimension you specify.

func NewSplitLayer added in v0.17.0 

func NewSplitLayer() *SplitLayer

NewSplitLayer creates a new SplitLayer.

func SplitLayerFromID added in v0.17.0 

func SplitLayerFromID(id objc.ID) *SplitLayer

SplitLayerFromID adopts an existing Objective-C object as a SplitLayer (nil for 0), retaining it and registering a release finalizer.

func (*SplitLayer) Dimension added in v0.17.0 

func (sl *SplitLayer) Dimension() int

Dimension returns the dimension (or axis) along which to split tensor

func (*SplitLayer) SplitCount added in v0.17.0 

func (sl *SplitLayer) SplitCount() int

SplitCount returns the number of splits. The tensor will be split into equally sized chunks. The last chunk may be smaller in size.

func (*SplitLayer) SplitSectionLengths added in v0.17.0 

func (sl *SplitLayer) SplitSectionLengths() []obj.Object

SplitSectionLengths returns lengths of each split section. The tensor will be split into chunks along dimensions with sizes given in \p splitSectionLengths .

SplitSectionLengths returns the collection as a Go slice.

func (*SplitLayer) WithIsDebuggingEnabled added in v0.17.0 

func (sl *SplitLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *SplitLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*SplitLayer) WithLabel added in v0.17.0 

func (sl *SplitLayer) WithLabel(label string) *SplitLayer

WithLabel sets a string that helps identify this layer.

type Tag added in v0.17.0 

type Tag int32
const (
	TagUndefinedTag  Tag = 0
	TagExtendedAllow Tag = 1
	TagExtendedDeny  Tag = 2
)

func (Tag) String added in v0.17.0 

func (e Tag) String() string

String returns the Tag constant's name, or its numeric form when the value is not a known constant.

type Tensor added in v0.17.0 

type Tensor struct {
	objref.Handle
}

Tensor is an idiomatic wrapper over the Objective-C class MLCTensor.

The data object you use throughout the framework.

func NewTensor added in v0.17.0 

func NewTensor() *Tensor

NewTensor creates a new Tensor.

func TensorFromID added in v0.17.0 

func TensorFromID(id objc.ID) *Tensor

TensorFromID adopts an existing Objective-C object as a Tensor (nil for 0), retaining it and registering a release finalizer.

func TensorWithDescriptor added in v0.17.0 

func TensorWithDescriptor(tensorDescriptor *TensorDescriptor) *Tensor

TensorWithDescriptor creates a tensor without data, using the descriptor you specify.

func TensorWithDescriptorData added in v0.17.0 

func TensorWithDescriptorData(tensorDescriptor *TensorDescriptor, data *TensorData) *Tensor

TensorWithDescriptorData creates a tensor with the descriptor and data you specify.

func TensorWithDescriptorFillWithData added in v0.17.0 

func TensorWithDescriptorFillWithData(tensorDescriptor *TensorDescriptor, fillData obj.Object) *Tensor

TensorWithDescriptorFillWithData creates a tensor with the descriptor and scalar value you specify.

func TensorWithDescriptorRandomInitializerType added in v0.17.0 

func TensorWithDescriptorRandomInitializerType(tensorDescriptor *TensorDescriptor, randomInitializerType RandomInitializerType) *Tensor

TensorWithDescriptorRandomInitializerType creates a tensor with the descriptor and random initializer type you specify.

func TensorWithSequenceLengthFeatureChannelCountBatchSize added in v0.17.0 

func TensorWithSequenceLengthFeatureChannelCountBatchSize(sequenceLength int, featureChannelCount int, batchSize int) *Tensor

TensorWithSequenceLengthFeatureChannelCountBatchSize creates a tensor without data, with the sequence length, number of feature channels, and batch size you specify.

func TensorWithSequenceLengthFeatureChannelCountBatchSizeData added in v0.17.0 

func TensorWithSequenceLengthFeatureChannelCountBatchSizeData(sequenceLength int, featureChannelCount int, batchSize int, data *TensorData) *Tensor

TensorWithSequenceLengthFeatureChannelCountBatchSizeData creates a tensor with the sequence length, number of feature channels, batch size, and data you specify.

func TensorWithSequenceLengthFeatureChannelCountBatchSizeRandomInitializerType added in v0.17.0 

func TensorWithSequenceLengthFeatureChannelCountBatchSizeRandomInitializerType(sequenceLength int, featureChannelCount int, batchSize int, randomInitializerType RandomInitializerType) *Tensor

TensorWithSequenceLengthFeatureChannelCountBatchSizeRandomInitializerType creates a tensor with the sequence length, number of feature channels, batch size, and random initializer type you specify.

func TensorWithSequenceLengthsSortedSequencesFeatureChannelCountBatchSizeData added in v0.17.0 

func TensorWithSequenceLengthsSortedSequencesFeatureChannelCountBatchSizeData(sequenceLengths []*foundation.Number, sortedSequences bool, featureChannelCount int, batchSize int, data *TensorData) *Tensor

TensorWithSequenceLengthsSortedSequencesFeatureChannelCountBatchSizeData creates a tensor with the sequence lengths, sorting indicator, number of feature channels, batch size, and data you specify.

func TensorWithSequenceLengthsSortedSequencesFeatureChannelCountBatchSizeRandomInitializerType added in v0.17.0 

func TensorWithSequenceLengthsSortedSequencesFeatureChannelCountBatchSizeRandomInitializerType(sequenceLengths []*foundation.Number, sortedSequences bool, featureChannelCount int, batchSize int, randomInitializerType RandomInitializerType) *Tensor

TensorWithSequenceLengthsSortedSequencesFeatureChannelCountBatchSizeRandomInitializerType creates a tensor with the sequence lengths, sorting indicator, number of feature channels, batch size, and random initializer type you specify.

func TensorWithShape added in v0.17.0 

func TensorWithShape(shape []*foundation.Number) *Tensor

TensorWithShape creates a tensor without data, with the shape you specify.

func TensorWithShapeDataDataType added in v0.17.0 

func TensorWithShapeDataDataType(shape []*foundation.Number, data *TensorData, dataType DataType) *Tensor

TensorWithShapeDataDataType creates a tensor with the shape, data, and data type you specify.

func TensorWithShapeDataType added in v0.17.0 

func TensorWithShapeDataType(shape []*foundation.Number, dataType DataType) *Tensor

TensorWithShapeDataType creates a tensor without data, with the shape and data type you specify.

func TensorWithShapeFillWithDataDataType added in v0.17.0 

func TensorWithShapeFillWithDataDataType(shape []*foundation.Number, fillData obj.Object, dataType DataType) *Tensor

TensorWithShapeFillWithDataDataType creates a tensor with the shape, scalar value, and data type you specify.

func TensorWithShapeRandomInitializerType added in v0.17.0 

func TensorWithShapeRandomInitializerType(shape []*foundation.Number, randomInitializerType RandomInitializerType) *Tensor

TensorWithShapeRandomInitializerType creates a tensor with the shape and random initializer type you specify.

func TensorWithShapeRandomInitializerTypeDataType added in v0.17.0 

func TensorWithShapeRandomInitializerTypeDataType(shape []*foundation.Number, randomInitializerType RandomInitializerType, dataType DataType) *Tensor

TensorWithShapeRandomInitializerTypeDataType creates a tensor with the shape, random initializer, and data type you specify.

func TensorWithWidthHeightFeatureChannelCountBatchSize added in v0.17.0 

func TensorWithWidthHeightFeatureChannelCountBatchSize(width int, height int, featureChannelCount int, batchSize int) *Tensor

TensorWithWidthHeightFeatureChannelCountBatchSize creates a tensor without data, with the sizes and number of feature channels you specify.

func TensorWithWidthHeightFeatureChannelCountBatchSizeData added in v0.17.0 

func TensorWithWidthHeightFeatureChannelCountBatchSizeData(width int, height int, featureChannelCount int, batchSize int, data *TensorData) *Tensor

TensorWithWidthHeightFeatureChannelCountBatchSizeData creates a tensor with the sizes, number of feature channels, and data you specify.

func TensorWithWidthHeightFeatureChannelCountBatchSizeDataDataType added in v0.17.0 

func TensorWithWidthHeightFeatureChannelCountBatchSizeDataDataType(width int, height int, featureChannelCount int, batchSize int, data *TensorData, dataType DataType) *Tensor

TensorWithWidthHeightFeatureChannelCountBatchSizeDataDataType creates a tensor with the sizes, number of feature channels, data, and data type you specify.

func TensorWithWidthHeightFeatureChannelCountBatchSizeFillWithDataDataType added in v0.17.0 

func TensorWithWidthHeightFeatureChannelCountBatchSizeFillWithDataDataType(width int, height int, featureChannelCount int, batchSize int, fillData float32, dataType DataType) *Tensor

TensorWithWidthHeightFeatureChannelCountBatchSizeFillWithDataDataType creates a tensor with the sizes and number of feature channels, and filled with the data and type you specify.

func TensorWithWidthHeightFeatureChannelCountBatchSizeRandomInitializerType added in v0.17.0 

func TensorWithWidthHeightFeatureChannelCountBatchSizeRandomInitializerType(width int, height int, featureChannelCount int, batchSize int, randomInitializerType RandomInitializerType) *Tensor

TensorWithWidthHeightFeatureChannelCountBatchSizeRandomInitializerType creates a tensor with the sizes, number of feature channels, and random data using the random initializer type you specify.

func (*Tensor) BindAndWriteDataToDevice added in v0.17.0 

func (t *Tensor) BindAndWriteDataToDevice(data *TensorData, device *Device) bool

BindAndWriteDataToDevice associates the given data to the tensor, and if the device is a GPU, also copies the data to the device memory.

func (*Tensor) BindOptimizerDataDeviceData added in v0.17.0 

func (t *Tensor) BindOptimizerDataDeviceData(data []*TensorData, deviceData []*TensorOptimizerDeviceData) bool

BindOptimizerDataDeviceData associates the optimizer and device data buffers you specify to the tensor.

func (*Tensor) CopyDataFromDeviceMemoryToBytesLengthSynchronizeWithDevice added in v0.17.0 

func (t *Tensor) CopyDataFromDeviceMemoryToBytesLengthSynchronizeWithDevice(data unsafe.Pointer, length int, synchronizeWithDevice bool) bool

CopyDataFromDeviceMemoryToBytesLengthSynchronizeWithDevice copies tensor data from device memory to user-specified memory.

func (*Tensor) Data added in v0.17.0 

func (t *Tensor) Data() []byte

Data returns the tensor data

func (*Tensor) Description added in v0.17.0 

func (t *Tensor) Description() string

Description returns the object's -description text.

func (*Tensor) Descriptor added in v0.17.0 

func (t *Tensor) Descriptor() *TensorDescriptor

Descriptor returns the tensor descriptor

func (*Tensor) Device added in v0.17.0 

func (t *Tensor) Device() *Device

Device returns the device associated with this tensor.

func (*Tensor) HasValidNumerics added in v0.17.0 

func (t *Tensor) HasValidNumerics() bool

HasValidNumerics reports whether returns a Boolean value indicating whether the underlying data has valid floating-point numerics, i.e. it does not contain NaN or INF floating-point values.

func (*Tensor) IsEqual added in v0.17.0 

func (t *Tensor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*Tensor) IsKind added in v0.17.0 

func (t *Tensor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*Tensor) Label added in v0.17.0 

func (t *Tensor) Label() string

Label returns a string to help identify this object.

func (*Tensor) OptimizerData added in v0.17.0 

func (t *Tensor) OptimizerData() []*TensorData

OptimizerData returns these are the host side optimizer (momentum and velocity) buffers which developers can query and initialize When customizing optimizer data, the contents of these buffers must be initialized before executing optimizer update for a graph.

OptimizerData returns the collection as a Go slice.

func (*Tensor) OptimizerDeviceData added in v0.17.0 

func (t *Tensor) OptimizerDeviceData() []*TensorOptimizerDeviceData

OptimizerDeviceData returns these are the device side optimizer (momentum and velocity) buffers which developers can query

OptimizerDeviceData returns the collection as a Go slice.

func (*Tensor) String added in v0.17.0 

func (t *Tensor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*Tensor) SynchronizeData added in v0.17.0 

func (t *Tensor) SynchronizeData() bool

SynchronizeData reports whether synchronizes the data in host memory.

func (*Tensor) SynchronizeOptimizerData added in v0.17.0 

func (t *Tensor) SynchronizeOptimizerData() bool

SynchronizeOptimizerData reports whether synchronizes the optimizer data in host memory.

func (*Tensor) TensorByDequantizingToTypeScaleBias added in v0.17.0 

func (t *Tensor) TensorByDequantizingToTypeScaleBias(type_ DataType, scale *Tensor, bias *Tensor) *Tensor

TensorByDequantizingToTypeScaleBias converts a tensor you quantize to a 32-bit floating-point tensor.

func (*Tensor) TensorByDequantizingToTypeScaleBiasAxis added in v0.17.0 

func (t *Tensor) TensorByDequantizingToTypeScaleBiasAxis(type_ DataType, scale *Tensor, bias *Tensor, axis int) *Tensor

TensorByDequantizingToTypeScaleBiasAxis converts a tensor you quantize to a 32-bit floating-point tensor.

func (*Tensor) TensorByQuantizingToTypeScaleBias added in v0.17.0 

func (t *Tensor) TensorByQuantizingToTypeScaleBias(type_ DataType, scale float32, bias int) *Tensor

TensorByQuantizingToTypeScaleBias converts a 32-bit floating-point tensor with the scale and bias you specify.

func (*Tensor) TensorByQuantizingToTypeScaleBiasAxis added in v0.17.0 

func (t *Tensor) TensorByQuantizingToTypeScaleBiasAxis(type_ DataType, scale *Tensor, bias *Tensor, axis int) *Tensor

TensorByQuantizingToTypeScaleBiasAxis converts a 32-bit floating-point tensor with the scale and bias you specify.

func (*Tensor) TensorID added in v0.17.0 

func (t *Tensor) TensorID() int

TensorID returns the tensor ID A unique number to identify each tensor. Assigned when the tensor is created.

func (*Tensor) WithLabel added in v0.17.0 

func (t *Tensor) WithLabel(label string) *Tensor

WithLabel sets a string that identifes this tensor.

type TensorData added in v0.17.0 

type TensorData struct {
	objref.Handle
}

TensorData is an idiomatic wrapper over the Objective-C class MLCTensorData.

An encapsulation of the memory that tensor data uses.

func DataWithBytesNoCopyLength added in v0.17.0 

func DataWithBytesNoCopyLength(data unsafe.Pointer, length int) *TensorData

DataWithBytesNoCopyLength creates a tensor data instance with the buffer of data and length of bytes you specify.

func DataWithBytesNoCopyLengthDeallocator added in v0.18.0 

func DataWithBytesNoCopyLengthDeallocator(data unsafe.Pointer, length int, deallocator func(unsafe.Pointer, int)) *TensorData

DataWithBytesNoCopyLengthDeallocator creates a tensor data instance with a data buffer, byte length, and custom deallocator closure you specify.

func DataWithImmutableBytesNoCopyLength added in v0.17.0 

func DataWithImmutableBytesNoCopyLength(data unsafe.Pointer, length int) *TensorData

DataWithImmutableBytesNoCopyLength creates a tensor data instance with the buffer of immutable data and length of bytes you specify.

func NewTensorData added in v0.17.0 

func NewTensorData() *TensorData

NewTensorData creates a new TensorData.

func TensorDataFromID added in v0.17.0 

func TensorDataFromID(id objc.ID) *TensorData

TensorDataFromID adopts an existing Objective-C object as a TensorData (nil for 0), retaining it and registering a release finalizer.

func (*TensorData) Bytes added in v0.18.0 

func (td *TensorData) Bytes() unsafe.Pointer

Bytes returns pointer to memory that contains or will be used for tensor data

func (*TensorData) Description added in v0.17.0 

func (td *TensorData) Description() string

Description returns the object's -description text.

func (*TensorData) IsEqual added in v0.17.0 

func (td *TensorData) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*TensorData) IsKind added in v0.17.0 

func (td *TensorData) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*TensorData) Length added in v0.17.0 

func (td *TensorData) Length() int

Length returns the size in bytes of the tensor data

func (*TensorData) String added in v0.17.0 

func (td *TensorData) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

type TensorDescriptor added in v0.17.0 

type TensorDescriptor struct {
	objref.Handle
}

TensorDescriptor is an idiomatic wrapper over the Objective-C class MLCTensorDescriptor.

A configuration object you use to create a tensor.

func ConvolutionBiasesDescriptorWithFeatureChannelCountDataType added in v0.17.0 

func ConvolutionBiasesDescriptorWithFeatureChannelCountDataType(featureChannelCount int, dataType DataType) *TensorDescriptor

ConvolutionBiasesDescriptorWithFeatureChannelCountDataType creates a tensor descriptor with the number of feature channels and data type you specify.

func ConvolutionWeightsDescriptorWithInputFeatureChannelCountOutputFeatureChannelCountDataType added in v0.17.0 

func ConvolutionWeightsDescriptorWithInputFeatureChannelCountOutputFeatureChannelCountDataType(inputFeatureChannelCount int, outputFeatureChannelCount int, dataType DataType) *TensorDescriptor

ConvolutionWeightsDescriptorWithInputFeatureChannelCountOutputFeatureChannelCountDataType creates a tensor descriptor with the number of feature channels and data type you specify.

func ConvolutionWeightsDescriptorWithWidthHeightInputFeatureChannelCountOutputFeatureChannelCountDataType added in v0.17.0 

func ConvolutionWeightsDescriptorWithWidthHeightInputFeatureChannelCountOutputFeatureChannelCountDataType(width int, height int, inputFeatureChannelCount int, outputFeatureChannelCount int, dataType DataType) *TensorDescriptor

ConvolutionWeightsDescriptorWithWidthHeightInputFeatureChannelCountOutputFeatureChannelCountDataType creates a tensor descriptor with the sizing, number of feature channels, and data type you specify.

func DescriptorWithShapeDataType added in v0.17.0 

func DescriptorWithShapeDataType(shape []*foundation.Number, dataType DataType) *TensorDescriptor

DescriptorWithShapeDataType creates a tensor descriptor with the shape and data type you specify.

func DescriptorWithShapeSequenceLengthsSortedSequencesDataType added in v0.17.0 

func DescriptorWithShapeSequenceLengthsSortedSequencesDataType(shape []*foundation.Number, sequenceLengths []*foundation.Number, sortedSequences bool, dataType DataType) *TensorDescriptor

DescriptorWithShapeSequenceLengthsSortedSequencesDataType creates a tensor descriptor with the shape, variable sequence lengths, sorting indicator, and data type you specify.

func DescriptorWithWidthHeightFeatureChannelCountBatchSize added in v0.17.0 

func DescriptorWithWidthHeightFeatureChannelCountBatchSize(width int, height int, featureChannels int, batchSize int) *TensorDescriptor

DescriptorWithWidthHeightFeatureChannelCountBatchSize creates a tensor descriptor with the width and height, number of feature channels, and batch size you specify.

func DescriptorWithWidthHeightFeatureChannelCountBatchSizeDataType added in v0.17.0 

func DescriptorWithWidthHeightFeatureChannelCountBatchSizeDataType(width int, height int, featureChannelCount int, batchSize int, dataType DataType) *TensorDescriptor

DescriptorWithWidthHeightFeatureChannelCountBatchSizeDataType creates a tensor descriptor with the width and height, number of feature channels, batch size, and data type you specify.

func NewTensorDescriptor added in v0.17.0 

func NewTensorDescriptor() *TensorDescriptor

NewTensorDescriptor creates a new TensorDescriptor.

func TensorDescriptorFromID added in v0.17.0 

func TensorDescriptorFromID(id objc.ID) *TensorDescriptor

TensorDescriptorFromID adopts an existing Objective-C object as a TensorDescriptor (nil for 0), retaining it and registering a release finalizer.

func (*TensorDescriptor) BatchSizePerSequenceStep added in v0.17.0 

func (td *TensorDescriptor) BatchSizePerSequenceStep() []obj.Object

BatchSizePerSequenceStep returns the batch size for each sequence We populate this only when sequenceLengths is valid. The length of this array should be the maximum sequence length in sequenceLengths (i.e sequenceLengths[0]).

BatchSizePerSequenceStep returns the collection as a Go slice.

func (*TensorDescriptor) DataType added in v0.17.0 

func (td *TensorDescriptor) DataType() DataType

DataType returns the tensor data type. The default is MLCDataTypeFloat32.

func (*TensorDescriptor) Description added in v0.17.0 

func (td *TensorDescriptor) Description() string

Description returns the object's -description text.

func (*TensorDescriptor) DimensionCount added in v0.17.0 

func (td *TensorDescriptor) DimensionCount() int

DimensionCount returns the number of dimensions in the tensor

func (*TensorDescriptor) IsEqual added in v0.17.0 

func (td *TensorDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*TensorDescriptor) IsKind added in v0.17.0 

func (td *TensorDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*TensorDescriptor) SequenceLengths added in v0.17.0 

func (td *TensorDescriptor) SequenceLengths() []obj.Object

SequenceLengths returns TODO

SequenceLengths returns the collection as a Go slice.

func (*TensorDescriptor) Shape added in v0.17.0 

func (td *TensorDescriptor) Shape() []obj.Object

Shape returns the size in each dimension

Shape returns the collection as a Go slice.

func (*TensorDescriptor) SortedSequences added in v0.17.0 

func (td *TensorDescriptor) SortedSequences() bool

SortedSequences reports whether the sequences are sorted or not.

func (*TensorDescriptor) Stride added in v0.17.0 

func (td *TensorDescriptor) Stride() []obj.Object

Stride returns the stride in bytes in each dimension

Stride returns the collection as a Go slice.

func (*TensorDescriptor) String added in v0.17.0 

func (td *TensorDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*TensorDescriptor) TensorAllocationSizeInBytes added in v0.17.0 

func (td *TensorDescriptor) TensorAllocationSizeInBytes() int

TensorAllocationSizeInBytes returns the allocation size in bytes for a tensor.

type TensorOptimizerDeviceData added in v0.17.0 

type TensorOptimizerDeviceData struct {
	objref.Handle
}

TensorOptimizerDeviceData is an idiomatic wrapper over the Objective-C class MLCTensorOptimizerDeviceData.

An encapsulation of the device memory associated with a tensor that an optimizer uses.

func NewTensorOptimizerDeviceData added in v0.17.0 

func NewTensorOptimizerDeviceData() *TensorOptimizerDeviceData

NewTensorOptimizerDeviceData creates a new TensorOptimizerDeviceData.

func TensorOptimizerDeviceDataFromID added in v0.17.0 

func TensorOptimizerDeviceDataFromID(id objc.ID) *TensorOptimizerDeviceData

TensorOptimizerDeviceDataFromID adopts an existing Objective-C object as a TensorOptimizerDeviceData (nil for 0), retaining it and registering a release finalizer.

func (*TensorOptimizerDeviceData) Description added in v0.17.0 

func (todd *TensorOptimizerDeviceData) Description() string

Description returns the object's -description text.

func (*TensorOptimizerDeviceData) IsEqual added in v0.17.0 

func (todd *TensorOptimizerDeviceData) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*TensorOptimizerDeviceData) IsKind added in v0.17.0 

func (todd *TensorOptimizerDeviceData) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*TensorOptimizerDeviceData) String added in v0.17.0 

func (todd *TensorOptimizerDeviceData) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

type TensorParameter added in v0.17.0 

type TensorParameter struct {
	objref.Handle
}

TensorParameter is an idiomatic wrapper over the Objective-C class MLCTensorParameter.

A tensor parameter object.

func NewTensorParameter added in v0.17.0 

func NewTensorParameter() *TensorParameter

NewTensorParameter creates a new TensorParameter.

func ParameterWithTensor added in v0.17.0 

func ParameterWithTensor(tensor *Tensor) *TensorParameter

ParameterWithTensor creates a tensor parameter with the tensor you specify.

func ParameterWithTensorOptimizerData added in v0.17.0 

func ParameterWithTensorOptimizerData(tensor *Tensor, optimizerData []*TensorData) *TensorParameter

ParameterWithTensorOptimizerData creates a tensor parameter with the tensor and optimizer data you specify.

func TensorParameterFromID added in v0.17.0 

func TensorParameterFromID(id objc.ID) *TensorParameter

TensorParameterFromID adopts an existing Objective-C object as a TensorParameter (nil for 0), retaining it and registering a release finalizer.

func (*TensorParameter) Description added in v0.17.0 

func (tp *TensorParameter) Description() string

Description returns the object's -description text.

func (*TensorParameter) IsEqual added in v0.17.0 

func (tp *TensorParameter) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*TensorParameter) IsKind added in v0.17.0 

func (tp *TensorParameter) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*TensorParameter) IsUpdatable added in v0.17.0 

func (tp *TensorParameter) IsUpdatable() bool

IsUpdatable reports whether this tensor parameter is updatable

func (*TensorParameter) String added in v0.17.0 

func (tp *TensorParameter) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*TensorParameter) Tensor added in v0.17.0 

func (tp *TensorParameter) Tensor() *Tensor

Tensor returns the underlying tensor

func (*TensorParameter) WithIsUpdatable added in v0.17.0 

func (tp *TensorParameter) WithIsUpdatable(isUpdatable bool) *TensorParameter

WithIsUpdatable sets a Boolean that indicates whether this tensor parameter is updatable.

type TrainingGraph added in v0.17.0 

type TrainingGraph struct {
	Graph
}

TrainingGraph is an idiomatic wrapper over the Objective-C class MLCTrainingGraph.

It embeds Graph, promoting that type's methods.

A training graph that you create from one or more graph objects plus additional layers you add directly to the training graph.

func GraphWithGraphObjectsLossLayerOptimizer added in v0.17.0 

func GraphWithGraphObjectsLossLayerOptimizer(graphObjects []*Graph, lossLayer *Layer, optimizer *Optimizer) *TrainingGraph

GraphWithGraphObjectsLossLayerOptimizer creates a training graph with the layers from the graph objects, loss layer, and optimizer you specify.

func NewTrainingGraph added in v0.17.0 

func NewTrainingGraph() *TrainingGraph

NewTrainingGraph creates a new TrainingGraph.

func TrainingGraphFromID added in v0.17.0 

func TrainingGraphFromID(id objc.ID) *TrainingGraph

TrainingGraphFromID adopts an existing Objective-C object as a TrainingGraph (nil for 0), retaining it and registering a release finalizer.

func (*TrainingGraph) AddInputsLossLabels added in v0.17.0 

func (tg *TrainingGraph) AddInputsLossLabels(inputs map[string]*Tensor, lossLabels map[string]*Tensor) bool

AddInputsLossLabels adds the inputs and loss label inputs that you specify to the training graph.

func (*TrainingGraph) AddInputsLossLabelsLossLabelWeights added in v0.17.0 

func (tg *TrainingGraph) AddInputsLossLabelsLossLabelWeights(inputs map[string]*Tensor, lossLabels map[string]*Tensor, lossLabelWeights map[string]*Tensor) bool

AddInputsLossLabelsLossLabelWeights adds the inputs, loss labels, and loss label weights that you specify to the training graph.

func (*TrainingGraph) AddOutputs added in v0.17.0 

func (tg *TrainingGraph) AddOutputs(outputs map[string]*Tensor) bool

AddOutputs adds the outputs to the training graph you specify.

func (*TrainingGraph) AllocateUserGradientForTensor added in v0.17.0 

func (tg *TrainingGraph) AllocateUserGradientForTensor(tensor *Tensor) *Tensor

AllocateUserGradientForTensor allocates an entry for a gradient for the result tensor you specify.

func (*TrainingGraph) BindOptimizerDataDeviceDataWithTensor added in v0.17.0 

func (tg *TrainingGraph) BindOptimizerDataDeviceDataWithTensor(data []*TensorData, deviceData []*TensorOptimizerDeviceData, tensor *Tensor) bool

BindOptimizerDataDeviceDataWithTensor associates the optimizer and device data you specify along with the tensor.

func (*TrainingGraph) CompileOptimizer added in v0.17.0 

func (tg *TrainingGraph) CompileOptimizer(optimizer *Optimizer) bool

CompileOptimizer compiles the optimizer to use with a training graph you specify.

func (*TrainingGraph) CompileWithOptionsDevice added in v0.17.0 

func (tg *TrainingGraph) CompileWithOptionsDevice(options GraphCompilationOptions, device *Device) bool

CompileWithOptionsDevice compiles the training graph for the options and device you specify.

func (*TrainingGraph) CompileWithOptionsDeviceInputTensorsInputTensorsData added in v0.17.0 

func (tg *TrainingGraph) CompileWithOptionsDeviceInputTensorsInputTensorsData(options GraphCompilationOptions, device *Device, inputTensors map[string]*Tensor, inputTensorsData map[string]*TensorData) bool

CompileWithOptionsDeviceInputTensorsInputTensorsData compiles the training graph for the options, device, and input tensors you specify.

func (*TrainingGraph) DeviceMemorySize added in v0.17.0 

func (tg *TrainingGraph) DeviceMemorySize() int

DeviceMemorySize returns the total size in bytes of device memory used for all intermediate tensors for forward, gradient passes and optimizer update for all layers in the training graph. We recommend executing an iteration before checking the device memory size as the buffers needed get allocated when the corresponding pass such as gradient, optimizer update is executed.

func (*TrainingGraph) ExecuteForwardWithBatchSizeOptionsCompletionHandler added in v0.18.0 

func (tg *TrainingGraph) ExecuteForwardWithBatchSizeOptionsCompletionHandler(batchSize int, options ExecutionOptions, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteForwardWithBatchSizeOptionsCompletionHandler executes the forward pass of the training graph with the batch size, execution options, and completion handler you specify.

func (*TrainingGraph) ExecuteForwardWithBatchSizeOptionsOutputsDataCompletionHandler added in v0.18.0 

func (tg *TrainingGraph) ExecuteForwardWithBatchSizeOptionsOutputsDataCompletionHandler(batchSize int, options ExecutionOptions, outputsData map[string]*TensorData, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteForwardWithBatchSizeOptionsOutputsDataCompletionHandler executes the forward pass of the training graph with the batch size, execution options, output data, and completion handler you specify.

func (*TrainingGraph) ExecuteGradientWithBatchSizeOptionsCompletionHandler added in v0.18.0 

func (tg *TrainingGraph) ExecuteGradientWithBatchSizeOptionsCompletionHandler(batchSize int, options ExecutionOptions, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteGradientWithBatchSizeOptionsCompletionHandler executes the gradient pass of the training graph with the batch size, execution options, and completion handler you specify.

func (*TrainingGraph) ExecuteGradientWithBatchSizeOptionsOutputsDataCompletionHandler added in v0.18.0 

func (tg *TrainingGraph) ExecuteGradientWithBatchSizeOptionsOutputsDataCompletionHandler(batchSize int, options ExecutionOptions, outputsData map[string]*TensorData, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteGradientWithBatchSizeOptionsOutputsDataCompletionHandler executes the gradient pass of the training graph with the batch size, execution options, output data, and completion handler you specify.

func (*TrainingGraph) ExecuteOptimizerUpdateWithOptionsCompletionHandler added in v0.18.0 

func (tg *TrainingGraph) ExecuteOptimizerUpdateWithOptionsCompletionHandler(options ExecutionOptions, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteOptimizerUpdateWithOptionsCompletionHandler executes the optimizer update pass of the training graph with the execution options and completion handler you specify.

func (*TrainingGraph) ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataBatchSizeOptionsCompletionHandler added in v0.18.0 

func (tg *TrainingGraph) ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataBatchSizeOptionsCompletionHandler(inputsData map[string]*TensorData, lossLabelsData map[string]*TensorData, lossLabelWeightsData map[string]*TensorData, batchSize int, options ExecutionOptions, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataBatchSizeOptionsCompletionHandler executes the training graph with the input data, batch size, execution options, and completion handler you specify.

func (*TrainingGraph) ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataOutputsDataBatchSizeOptionsCompletionHandler added in v0.18.0 

func (tg *TrainingGraph) ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataOutputsDataBatchSizeOptionsCompletionHandler(inputsData map[string]*TensorData, lossLabelsData map[string]*TensorData, lossLabelWeightsData map[string]*TensorData, outputsData map[string]*TensorData, batchSize int, options ExecutionOptions, completionHandler func(obj.Object, unsafe.Pointer, float64)) bool

ExecuteWithInputsDataLossLabelsDataLossLabelWeightsDataOutputsDataBatchSizeOptionsCompletionHandler executes the training graph with the input data, output data, batch size, execution options, and completion handler that you specify.

func (*TrainingGraph) GradientDataForParameterLayer added in v0.17.0 

func (tg *TrainingGraph) GradientDataForParameterLayer(parameter *Tensor, layer *Layer) []byte

GradientDataForParameterLayer gets the gradient data for the trainable parameter and associated layer you specify.

func (*TrainingGraph) GradientTensorForInput added in v0.17.0 

func (tg *TrainingGraph) GradientTensorForInput(input *Tensor) *Tensor

GradientTensorForInput gets the gradient tensor for the input tensor you specify.

func (*TrainingGraph) LinkWithGraphs added in v0.17.0 

func (tg *TrainingGraph) LinkWithGraphs(graphs []*TrainingGraph) bool

LinkWithGraphs links the training graphs you specify.

func (*TrainingGraph) Optimizer added in v0.17.0 

func (tg *TrainingGraph) Optimizer() *Optimizer

Optimizer returns the optimizer to be used with the training graph

func (*TrainingGraph) ResultGradientTensorsForLayer added in v0.17.0 

func (tg *TrainingGraph) ResultGradientTensorsForLayer(layer *Layer) []*Tensor

ResultGradientTensorsForLayer gets the result gradient tensors for the layer in the training graph you specify.

func (*TrainingGraph) SetTrainingTensorParameters added in v0.17.0 

func (tg *TrainingGraph) SetTrainingTensorParameters(parameters []*TensorParameter) bool

SetTrainingTensorParameters sets the input tensor parameters, which the optimizer then updates.

func (*TrainingGraph) SourceGradientTensorsForLayer added in v0.17.0 

func (tg *TrainingGraph) SourceGradientTensorsForLayer(layer *Layer) []*Tensor

SourceGradientTensorsForLayer gets the source gradient tensors for the layer in the training graph you specify.

func (*TrainingGraph) StopGradientForTensors added in v0.17.0 

func (tg *TrainingGraph) StopGradientForTensors(tensors []*Tensor) bool

StopGradientForTensors adds the tensors that you specify, to indicate which contributions the graph excludes when computing gradients during gradient pass.

func (*TrainingGraph) SynchronizeUpdates added in v0.17.0 

func (tg *TrainingGraph) SynchronizeUpdates()

SynchronizeUpdates synchronizes updates from device memory.

type TransposeLayer added in v0.17.0 

type TransposeLayer struct {
	Layer
}

TransposeLayer is an idiomatic wrapper over the Objective-C class MLCTransposeLayer.

It embeds Layer, promoting that type's methods.

A layer that permutes the dimensions you specify.

func LayerWithDimensions added in v0.17.0 

func LayerWithDimensions(dimensions []*foundation.Number) *TransposeLayer

LayerWithDimensions creates a transpose layer with the dimensions you specify.

func NewTransposeLayer added in v0.17.0 

func NewTransposeLayer() *TransposeLayer

NewTransposeLayer creates a new TransposeLayer.

func TransposeLayerFromID added in v0.17.0 

func TransposeLayerFromID(id objc.ID) *TransposeLayer

TransposeLayerFromID adopts an existing Objective-C object as a TransposeLayer (nil for 0), retaining it and registering a release finalizer.

func (*TransposeLayer) Dimensions added in v0.17.0 

func (tl *TransposeLayer) Dimensions() []obj.Object

Dimensions returns permutes the dimensions according to 'dimensions'. The returned tensor's dimension i will correspond to dimensions[i].

Dimensions returns the collection as a Go slice.

func (*TransposeLayer) WithIsDebuggingEnabled added in v0.17.0 

func (tl *TransposeLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *TransposeLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*TransposeLayer) WithLabel added in v0.17.0 

func (tl *TransposeLayer) WithLabel(label string) *TransposeLayer

WithLabel sets a string that helps identify this layer.

type Type added in v0.17.0 

type Type int32
const (
	TypeExtended Type = 256
	TypeAccess   Type = 0
	TypeDefault  Type = 1
	TypeAfs      Type = 2
	TypeCoda     Type = 3
	TypeNtfs     Type = 4
	TypeNwfs     Type = 5
)

func (Type) String added in v0.17.0 

func (e Type) String() string

String returns the Type constant's name, or its numeric form when the value is not a known constant.

type UpsampleLayer added in v0.17.0 

type UpsampleLayer struct {
	Layer
}

UpsampleLayer is an idiomatic wrapper over the Objective-C class MLCUpsampleLayer.

It embeds Layer, promoting that type's methods.

A layer that applies upsampling with the shape you specify.

func LayerWithShapeSampleModeAlignsCorners added in v0.17.0 

func LayerWithShapeSampleModeAlignsCorners(shape []*foundation.Number, sampleMode SampleMode, alignsCorners bool) *UpsampleLayer

LayerWithShapeSampleModeAlignsCorners creates an upsample layer with the shape, upsampling algorithm, and corner alignement option you specify.

func MLCUpsampleLayerLayerWithShape 

func MLCUpsampleLayerLayerWithShape(shape []*foundation.Number) *UpsampleLayer

MLCUpsampleLayerLayerWithShape creates an upsample layer with the shape you specify.

func NewUpsampleLayer added in v0.17.0 

func NewUpsampleLayer() *UpsampleLayer

NewUpsampleLayer creates a new UpsampleLayer.

func UpsampleLayerFromID added in v0.17.0 

func UpsampleLayerFromID(id objc.ID) *UpsampleLayer

UpsampleLayerFromID adopts an existing Objective-C object as a UpsampleLayer (nil for 0), retaining it and registering a release finalizer.

func (*UpsampleLayer) AlignsCorners added in v0.17.0 

func (ul *UpsampleLayer) AlignsCorners() bool

AlignsCorners reports whether a boolean that specifies whether the corner pixels of the source and result tensors are aligned. If True, the corner pixels of the source and result tensors are aligned, and thus preserving the values at those pixels. This only has effect when mode is 'bilinear'. Default is false.

func (*UpsampleLayer) SampleMode added in v0.17.0 

func (ul *UpsampleLayer) SampleMode() SampleMode

SampleMode returns the sampling mode to use when performing the upsample.

func (*UpsampleLayer) Shape added in v0.17.0 

func (ul *UpsampleLayer) Shape() []obj.Object

Shape returns a NSArray<NSNumber *> representing just the width if number of entries in shape array is 1 or the height followed by width of result tensor if the number of entries in shape array is 2.

Shape returns the collection as a Go slice.

func (*UpsampleLayer) WithIsDebuggingEnabled added in v0.17.0 

func (ul *UpsampleLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *UpsampleLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*UpsampleLayer) WithLabel added in v0.17.0 

func (ul *UpsampleLayer) WithLabel(label string) *UpsampleLayer

WithLabel sets a string that helps identify this layer.

type VirtualMemoryGuardExceptionCode added in v0.17.0 

type VirtualMemoryGuardExceptionCode uint32
const (
	KGUARD_EXC_DEALLOC_GAP                   VirtualMemoryGuardExceptionCode = 1
	KGUARD_EXC_RECLAIM_COPYIO_FAILURE        VirtualMemoryGuardExceptionCode = 2
	KGUARD_EXC_RECLAIM_INDEX_FAILURE         VirtualMemoryGuardExceptionCode = 4
	KGUARD_EXC_RECLAIM_DEALLOCATE_FAILURE    VirtualMemoryGuardExceptionCode = 8
	KGUARD_EXC_RECLAIM_ACCOUNTING_FAILURE    VirtualMemoryGuardExceptionCode = 9
	KGUARD_EXC_SEC_IOPL_ON_EXEC_PAGE         VirtualMemoryGuardExceptionCode = 10
	KGUARD_EXC_SEC_EXEC_ON_IOPL_PAGE         VirtualMemoryGuardExceptionCode = 11
	KGUARD_EXC_SEC_UPL_WRITE_ON_EXEC_REGION  VirtualMemoryGuardExceptionCode = 12
	KGUARD_EXC_LARGE_ALLOCATION_TELEMETRY    VirtualMemoryGuardExceptionCode = 13
	KGUARD_EXC_SEC_ACCESS_FAULT              VirtualMemoryGuardExceptionCode = 98
	KGUARD_EXC_SEC_ASYNC_ACCESS_FAULT        VirtualMemoryGuardExceptionCode = 99
	KGUARD_EXC_SEC_COPY_DENIED               VirtualMemoryGuardExceptionCode = 100
	KGUARD_EXC_SEC_SHARING_DENIED            VirtualMemoryGuardExceptionCode = 101
	KGUARD_EXC_MTE_SYNC_FAULT                VirtualMemoryGuardExceptionCode = 200
	KGUARD_EXC_MTE_ASYNC_USER_FAULT          VirtualMemoryGuardExceptionCode = 201
	KGUARD_EXC_MTE_ASYNC_KERN_FAULT          VirtualMemoryGuardExceptionCode = 202
	KGUARD_EXC_GUARD_OBJECT_ASYNC_USER_FAULT VirtualMemoryGuardExceptionCode = 203
	KGUARD_EXC_GUARD_OBJECT_ASYNC_KERN_FAULT VirtualMemoryGuardExceptionCode = 204
)

func (VirtualMemoryGuardExceptionCode) String added in v0.17.0 

func (e VirtualMemoryGuardExceptionCode) String() string

String returns the VirtualMemoryGuardExceptionCode constant's name, or its numeric form when the value is not a known constant.

type XpcListenerCreateFlags added in v0.17.0 

type XpcListenerCreateFlags uint64

Bitmask — values may be combined with |. 

const (
	XpcListenerCreateFlagsNone            XpcListenerCreateFlags = 0
	XpcListenerCreateFlagsInactive        XpcListenerCreateFlags = 1
	XpcListenerCreateFlagsForceMach       XpcListenerCreateFlags = 2
	XpcListenerCreateFlagsForceXpcservice XpcListenerCreateFlags = 4
)

func (XpcListenerCreateFlags) String added in v0.17.0 

func (e XpcListenerCreateFlags) String() string

String returns the XpcListenerCreateFlags constant's name, or its numeric form when the value is not a known constant.

type XpcSessionCreateFlags added in v0.17.0 

type XpcSessionCreateFlags uint64

Bitmask — values may be combined with |. 

const (
	XpcSessionCreateFlagsNone           XpcSessionCreateFlags = 0
	XpcSessionCreateFlagsInactive       XpcSessionCreateFlags = 1
	XpcSessionCreateFlagsMachPrivileged XpcSessionCreateFlags = 2
)

func (XpcSessionCreateFlags) String added in v0.17.0 

func (e XpcSessionCreateFlags) String() string

String returns the XpcSessionCreateFlags constant's name, or its numeric form when the value is not a known constant.

type YOLOLossDescriptor added in v0.17.0 

type YOLOLossDescriptor struct {
	objref.Handle
}

YOLOLossDescriptor is an idiomatic wrapper over the Objective-C class MLCYOLOLossDescriptor.

The configuration object you use to create the YOLO loss layer.

func DescriptorWithAnchorBoxesAnchorBoxCount added in v0.17.0 

func DescriptorWithAnchorBoxesAnchorBoxCount(anchorBoxes []byte, anchorBoxCount int) *YOLOLossDescriptor

DescriptorWithAnchorBoxesAnchorBoxCount creates a YOLO loss filter descriptor with the anchor box data and number of anchor boxes you specify.

func NewYOLOLossDescriptor added in v0.17.0 

func NewYOLOLossDescriptor() *YOLOLossDescriptor

NewYOLOLossDescriptor creates a new YOLOLossDescriptor.

func YOLOLossDescriptorFromID added in v0.17.0 

func YOLOLossDescriptorFromID(id objc.ID) *YOLOLossDescriptor

YOLOLossDescriptorFromID adopts an existing Objective-C object as a YOLOLossDescriptor (nil for 0), retaining it and registering a release finalizer.

func (*YOLOLossDescriptor) AnchorBoxCount added in v0.17.0 

func (yld *YOLOLossDescriptor) AnchorBoxCount() int

AnchorBoxCount returns number of anchor boxes used to detect object per grid cell

func (*YOLOLossDescriptor) AnchorBoxes added in v0.17.0 

func (yld *YOLOLossDescriptor) AnchorBoxes() []byte

AnchorBoxes returns \p NSData containing the width and height for \p anchorBoxCount anchor boxes This \p NSData should have 2 floating-point values per anchor box which represent the width and height of the anchor box.

func (*YOLOLossDescriptor) Description added in v0.17.0 

func (yld *YOLOLossDescriptor) Description() string

Description returns the object's -description text.

func (*YOLOLossDescriptor) IsEqual added in v0.17.0 

func (yld *YOLOLossDescriptor) IsEqual(other obj.Object) bool

IsEqual reports Objective-C equality (isEqual:) with another object.

func (*YOLOLossDescriptor) IsKind added in v0.17.0 

func (yld *YOLOLossDescriptor) IsKind(className string) bool

IsKind reports whether the object is an instance of the named class or a subclass.

func (*YOLOLossDescriptor) MaximumIOUForObjectAbsence added in v0.17.0 

func (yld *YOLOLossDescriptor) MaximumIOUForObjectAbsence() float32

MaximumIOUForObjectAbsence returns if the prediction IOU with groundTruth is lower than this value we consider it a confident object absence. The default is 0.3

func (*YOLOLossDescriptor) MinimumIOUForObjectPresence added in v0.17.0 

func (yld *YOLOLossDescriptor) MinimumIOUForObjectPresence() float32

MinimumIOUForObjectPresence returns if the prediction IOU with groundTruth is higher than this value we consider it a confident object presence, The default is 0.7

func (*YOLOLossDescriptor) ScaleClassLoss added in v0.17.0 

func (yld *YOLOLossDescriptor) ScaleClassLoss() float32

ScaleClassLoss returns the scale factor for no object classes loss and loss gradient. The default is 2.0

func (*YOLOLossDescriptor) ScaleNoObjectConfidenceLoss added in v0.17.0 

func (yld *YOLOLossDescriptor) ScaleNoObjectConfidenceLoss() float32

ScaleNoObjectConfidenceLoss returns the scale factor for no object confidence loss and loss gradient. The default is 5.0

func (*YOLOLossDescriptor) ScaleObjectConfidenceLoss added in v0.17.0 

func (yld *YOLOLossDescriptor) ScaleObjectConfidenceLoss() float32

ScaleObjectConfidenceLoss returns the scale factor for object confidence loss and loss gradient. The default is 100.0

func (*YOLOLossDescriptor) ScaleSpatialPositionLoss added in v0.17.0 

func (yld *YOLOLossDescriptor) ScaleSpatialPositionLoss() float32

ScaleSpatialPositionLoss returns the scale factor for spatial position loss and loss gradient. The default is 10.0

func (*YOLOLossDescriptor) ScaleSpatialSizeLoss added in v0.17.0 

func (yld *YOLOLossDescriptor) ScaleSpatialSizeLoss() float32

ScaleSpatialSizeLoss returns the scale factor for spatial size loss and loss gradient. The default is 10.0

func (*YOLOLossDescriptor) ShouldRescore added in v0.17.0 

func (yld *YOLOLossDescriptor) ShouldRescore() bool

ShouldRescore reports whether rescore pertains to multiplying the confidence groundTruth with IOU (intersection over union) of predicted bounding box and the groundTruth boundingBox. The default is true

func (*YOLOLossDescriptor) String added in v0.17.0 

func (yld *YOLOLossDescriptor) String() string

String returns the object's -description text, so a wrapper prints usefully under fmt.

func (*YOLOLossDescriptor) WithMaximumIOUForObjectAbsence added in v0.17.0 

func (yld *YOLOLossDescriptor) WithMaximumIOUForObjectAbsence(maximumIOUForObjectAbsence float32) *YOLOLossDescriptor

WithMaximumIOUForObjectAbsence sets the negative intersection over union (IOU).

func (*YOLOLossDescriptor) WithMinimumIOUForObjectPresence added in v0.17.0 

func (yld *YOLOLossDescriptor) WithMinimumIOUForObjectPresence(minimumIOUForObjectPresence float32) *YOLOLossDescriptor

WithMinimumIOUForObjectPresence sets the positive intersection over union (IOU).

func (*YOLOLossDescriptor) WithScaleClassLoss added in v0.17.0 

func (yld *YOLOLossDescriptor) WithScaleClassLoss(scaleClassLoss float32) *YOLOLossDescriptor

WithScaleClassLoss sets the scale factor you use for loss when there are no object classes, and for loss gradient.

func (*YOLOLossDescriptor) WithScaleNoObjectConfidenceLoss added in v0.17.0 

func (yld *YOLOLossDescriptor) WithScaleNoObjectConfidenceLoss(scaleNoObjectConfidenceLoss float32) *YOLOLossDescriptor

WithScaleNoObjectConfidenceLoss sets the scale factor you use for no object confidence loss and loss gradient.

func (*YOLOLossDescriptor) WithScaleObjectConfidenceLoss added in v0.17.0 

func (yld *YOLOLossDescriptor) WithScaleObjectConfidenceLoss(scaleObjectConfidenceLoss float32) *YOLOLossDescriptor

WithScaleObjectConfidenceLoss sets the scale factor you use for object confidence loss and loss gradient.

func (*YOLOLossDescriptor) WithScaleSpatialPositionLoss added in v0.17.0 

func (yld *YOLOLossDescriptor) WithScaleSpatialPositionLoss(scaleSpatialPositionLoss float32) *YOLOLossDescriptor

WithScaleSpatialPositionLoss sets the scale factor you use for spatial position loss and loss gradient.

func (*YOLOLossDescriptor) WithScaleSpatialSizeLoss added in v0.17.0 

func (yld *YOLOLossDescriptor) WithScaleSpatialSizeLoss(scaleSpatialSizeLoss float32) *YOLOLossDescriptor

WithScaleSpatialSizeLoss sets the scale factor you use for spatial size loss and loss gradient.

func (*YOLOLossDescriptor) WithShouldRescore added in v0.17.0 

func (yld *YOLOLossDescriptor) WithShouldRescore(shouldRescore bool) *YOLOLossDescriptor

WithShouldRescore sets a Boolean that indicates whether the layer scales the object confidence loss by the intersection over union (IOU) overlap.

type YOLOLossLayer added in v0.17.0 

type YOLOLossLayer struct {
	LossLayer
}

YOLOLossLayer is an idiomatic wrapper over the Objective-C class MLCYOLOLossLayer.

It embeds LossLayer, promoting that type's methods.

A layer that estimates loss for the YOLO algorithm.

func MLCYOLOLossLayerLayerWithDescriptor 

func MLCYOLOLossLayerLayerWithDescriptor(lossDescriptor *YOLOLossDescriptor) *YOLOLossLayer

MLCYOLOLossLayerLayerWithDescriptor creates a YOLO loss layer with the descriptor you specify.

func NewYOLOLossLayer added in v0.17.0 

func NewYOLOLossLayer() *YOLOLossLayer

NewYOLOLossLayer creates a new YOLOLossLayer.

func YOLOLossLayerFromID added in v0.17.0 

func YOLOLossLayerFromID(id objc.ID) *YOLOLossLayer

YOLOLossLayerFromID adopts an existing Objective-C object as a YOLOLossLayer (nil for 0), retaining it and registering a release finalizer.

func (*YOLOLossLayer) WithIsDebuggingEnabled added in v0.17.0 

func (yll *YOLOLossLayer) WithIsDebuggingEnabled(isDebuggingEnabled bool) *YOLOLossLayer

WithIsDebuggingEnabled sets a Boolean that indicates whether you choose to debug the layer when executing a graph that includes it.

func (*YOLOLossLayer) WithLabel added in v0.17.0 

func (yll *YOLOLossLayer) WithLabel(label string) *YOLOLossLayer

WithLabel sets a string that helps identify this layer.

func (*YOLOLossLayer) YoloLossDescriptor added in v0.17.0 

func (yll *YOLOLossLayer) YoloLossDescriptor() *YOLOLossDescriptor

YoloLossDescriptor returns the YOLO loss descriptor

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.