objc2-metal-performance-shaders 0.3.2

//! This file has been automatically generated by `objc2`'s `header-translator`.
//! DO NOT EDIT
use core::ffi::*;
use core::ptr::NonNull;
use objc2::__framework_prelude::*;
use objc2_foundation::*;
use objc2_metal::*;

use crate::*;

/// [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnnsequencedirection?language=objc)
// NS_ENUM
#[repr(transparent)]
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, PartialOrd, Ord)]
pub struct MPSRNNSequenceDirection(pub NSUInteger);
impl MPSRNNSequenceDirection {
    /// The input sequence is processed from index zero to array length minus one
    #[doc(alias = "MPSRNNSequenceDirectionForward")]
    pub const Forward: Self = Self(0);
    /// The input sequence is processed from index array length minus one to zero
    #[doc(alias = "MPSRNNSequenceDirectionBackward")]
    pub const Backward: Self = Self(1);
}

unsafe impl Encode for MPSRNNSequenceDirection {
    const ENCODING: Encoding = NSUInteger::ENCODING;
}

unsafe impl RefEncode for MPSRNNSequenceDirection {
    const ENCODING_REF: Encoding = Encoding::Pointer(&Self::ENCODING);
}

/// [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnnbidirectionalcombinemode?language=objc)
// NS_ENUM
#[repr(transparent)]
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, PartialOrd, Ord)]
pub struct MPSRNNBidirectionalCombineMode(pub NSUInteger);
impl MPSRNNBidirectionalCombineMode {
    /// The two sequences are kept separate
    #[doc(alias = "MPSRNNBidirectionalCombineModeNone")]
    pub const None: Self = Self(0);
    /// The two sequences are summed together to form a single output
    #[doc(alias = "MPSRNNBidirectionalCombineModeAdd")]
    pub const Add: Self = Self(1);
    /// The two sequences are concatenated together along the feature channels to form a single output
    #[doc(alias = "MPSRNNBidirectionalCombineModeConcatenate")]
    pub const Concatenate: Self = Self(2);
}

unsafe impl Encode for MPSRNNBidirectionalCombineMode {
    const ENCODING: Encoding = NSUInteger::ENCODING;
}

unsafe impl RefEncode for MPSRNNBidirectionalCombineMode {
    const ENCODING_REF: Encoding = Encoding::Pointer(&Self::ENCODING);
}

extern_class!(
    /// Dependencies: This depends on Metal.framework
    ///
    /// The MPSRNNDescriptor specifies a Recursive neural network block/layer descriptor.
    ///
    /// See also [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnndescriptor?language=objc)
    #[unsafe(super(NSObject))]
    #[derive(Debug, PartialEq, Eq, Hash)]
    pub struct MPSRNNDescriptor;
);

extern_conformance!(
    unsafe impl NSObjectProtocol for MPSRNNDescriptor {}
);

impl MPSRNNDescriptor {
    extern_methods!(
        /// The number of feature channels per pixel in the input image or number of rows in the input matrix.
        #[unsafe(method(inputFeatureChannels))]
        #[unsafe(method_family = none)]
        pub unsafe fn inputFeatureChannels(&self) -> NSUInteger;

        /// Setter for [`inputFeatureChannels`][Self::inputFeatureChannels].
        #[unsafe(method(setInputFeatureChannels:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setInputFeatureChannels(&self, input_feature_channels: NSUInteger);

        /// The number of feature channels per pixel in the destination image or number of rows in the destination matrix.
        #[unsafe(method(outputFeatureChannels))]
        #[unsafe(method_family = none)]
        pub unsafe fn outputFeatureChannels(&self) -> NSUInteger;

        /// Setter for [`outputFeatureChannels`][Self::outputFeatureChannels].
        #[unsafe(method(setOutputFeatureChannels:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setOutputFeatureChannels(&self, output_feature_channels: NSUInteger);

        /// if YES then use identity transformation for all weights (W, Wr, Wi, Wf, Wo, Wc) affecting input x_j in this layer,
        /// even if said weights are specified as nil.
        /// For example 'W_ij * x_j' is replaced by 'x_j' in formulae defined in
        /// MPSRNNSingleGateDescriptor.Defaults to NO.
        #[unsafe(method(useLayerInputUnitTransformMode))]
        #[unsafe(method_family = none)]
        pub unsafe fn useLayerInputUnitTransformMode(&self) -> bool;

        /// Setter for [`useLayerInputUnitTransformMode`][Self::useLayerInputUnitTransformMode].
        #[unsafe(method(setUseLayerInputUnitTransformMode:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setUseLayerInputUnitTransformMode(
            &self,
            use_layer_input_unit_transform_mode: bool,
        );

        /// If YES, then
        /// MPSRNNMatrixInferenceLayeruses 32-bit floating point numbers internally for weights when
        /// computing matrix transformations. If NO, then 16-bit, half precision floating point numbers are used.
        /// Currently
        /// MPSRNNImageInferenceLayerignores this property and the convolution operations always
        /// convert FP32 weights into FP16 for better performance.
        /// Defaults to NO.
        #[unsafe(method(useFloat32Weights))]
        #[unsafe(method_family = none)]
        pub unsafe fn useFloat32Weights(&self) -> bool;

        /// Setter for [`useFloat32Weights`][Self::useFloat32Weights].
        #[unsafe(method(setUseFloat32Weights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setUseFloat32Weights(&self, use_float32_weights: bool);

        /// When the layer specified with this descriptor is used to process a sequence of inputs
        /// by calling
        ///
        /// See: encodeBidirectionalSequenceToCommandBuffer then this parameter defines
        /// in which direction the sequence is processed. The operation of the layer is:
        /// (yt, ht, ct) = f(xt,ht-1,ct-1) for MPSRNNSequenceDirectionForward
        /// and
        /// (yt, ht, ct) = f(xt,ht+1,ct+1) for MPSRNNSequenceDirectionBackward, where
        /// xt is the output of the previous layer that encodes in the same direction as this layer,
        /// (or the input image or matrix if this is the first layer in stack with this direction).
        ///
        /// See: MPSRNNImageInferenceLayer and
        ///
        /// See: MPSRNNMatrixInferenceLayer.
        #[unsafe(method(layerSequenceDirection))]
        #[unsafe(method_family = none)]
        pub unsafe fn layerSequenceDirection(&self) -> MPSRNNSequenceDirection;

        /// Setter for [`layerSequenceDirection`][Self::layerSequenceDirection].
        #[unsafe(method(setLayerSequenceDirection:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setLayerSequenceDirection(
            &self,
            layer_sequence_direction: MPSRNNSequenceDirection,
        );
    );
}

/// Methods declared on superclass `NSObject`.
impl MPSRNNDescriptor {
    extern_methods!(
        #[unsafe(method(init))]
        #[unsafe(method_family = init)]
        pub unsafe fn init(this: Allocated<Self>) -> Retained<Self>;

        #[unsafe(method(new))]
        #[unsafe(method_family = new)]
        pub unsafe fn new() -> Retained<Self>;
    );
}

extern_class!(
    /// Dependencies: This depends on Metal.framework
    ///
    /// The MPSRNNSingleGateDescriptor specifies a simple recurrent block/layer descriptor.
    /// The RNN layer initialized with a MPSRNNSingleGateDescriptor transforms the input data (image or matrix),
    /// and previous output with a set of filters, each producing one feature map in the new output data.
    /// The user may provide the RNN unit a single input or a sequence of inputs.
    ///
    /// Description of operation:
    ///
    /// Let x_j be the input data (at time index t of sequence,
    /// j index containing quadruplet: batch index, x,y and feature index (x=y=0 for matrices)).
    /// Let h0_j be the recurrent input (previous output) data from previous time step (at time index t-1 of sequence).
    /// Let h1_i be the output data produced at this time step.
    ///
    /// Let W_ij, U_ij be the weights for input and recurrent input data respectively
    /// Let b_i be a bias term
    ///
    /// Let gi(x) be a neuron activation function
    ///
    /// Then the new output image h1_i data is computed as follows:
    ///
    /// h1_i = gi( W_ij * x_j + U_ij * h0_j  + b_i )
    ///
    /// The '*' stands for convolution (see
    /// MPSRNNImageInferenceLayer)or matrix-vector/matrix multiplication
    /// (see
    /// MPSRNNMatrixInferenceLayer).Summation is over index j (except for the batch index), but there is no summation over
    /// repeated index i - the output index.
    /// Note that for validity all intermediate images have to be of same size and the U matrix has to be square
    /// (ie. outputFeatureChannels == inputFeatureChannels in those). Also the bias terms are scalars wrt. spatial dimensions.
    ///
    /// See also [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnnsinglegatedescriptor?language=objc)
    #[unsafe(super(MPSRNNDescriptor, NSObject))]
    #[derive(Debug, PartialEq, Eq, Hash)]
    pub struct MPSRNNSingleGateDescriptor;
);

extern_conformance!(
    unsafe impl NSObjectProtocol for MPSRNNSingleGateDescriptor {}
);

impl MPSRNNSingleGateDescriptor {
    extern_methods!(
        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'W_ij', bias 'b_i' and neuron 'gi' from the simple RNN layer formula.
        /// If nil then assumed zero weights, bias and no neuron (identity mapping). Defaults to nil.
        #[unsafe(method(inputWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn inputWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`inputWeights`][Self::inputWeights].
        #[unsafe(method(setInputWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setInputWeights(
            &self,
            input_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'U_ij' from the simple RNN layer formula.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(recurrentWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn recurrentWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`recurrentWeights`][Self::recurrentWeights].
        #[unsafe(method(setRecurrentWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setRecurrentWeights(
            &self,
            recurrent_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        /// Creates a MPSRNNSingleGateDescriptor
        ///
        /// Parameter `inputFeatureChannels`: The number of feature channels in the input image/matrix. Must be >= 1.
        ///
        /// Parameter `outputFeatureChannels`: The number of feature channels in the output image/matrix. Must be >= 1.
        ///
        /// Returns: A valid MPSRNNSingleGateDescriptor object or nil, if failure.
        #[unsafe(method(createRNNSingleGateDescriptorWithInputFeatureChannels:outputFeatureChannels:))]
        #[unsafe(method_family = none)]
        pub unsafe fn createRNNSingleGateDescriptorWithInputFeatureChannels_outputFeatureChannels(
            input_feature_channels: NSUInteger,
            output_feature_channels: NSUInteger,
        ) -> Retained<Self>;
    );
}

/// Methods declared on superclass `NSObject`.
impl MPSRNNSingleGateDescriptor {
    extern_methods!(
        #[unsafe(method(init))]
        #[unsafe(method_family = init)]
        pub unsafe fn init(this: Allocated<Self>) -> Retained<Self>;

        #[unsafe(method(new))]
        #[unsafe(method_family = new)]
        pub unsafe fn new() -> Retained<Self>;
    );
}

extern_class!(
    /// Dependencies: This depends on Metal.framework
    ///
    /// The MPSGRUDescriptor specifies a GRU (Gated Recurrent Unit) block/layer descriptor.
    /// The RNN layer initialized with a MPSGRUDescriptor transforms the input data (image or matrix),
    /// and previous output with a set of filters, each producing one feature map in
    /// the output data according to the Gated unit formulae detailed below.
    /// The user may provide the GRU unit a single input or a sequence of inputs. The layer also supports
    /// p-norm gating (Detailed in: https://arxiv.org/abs/1608.03639 ).
    ///
    /// Description of operation:
    ///
    /// Let x_j be the input data (at time index t of sequence,
    /// j index containing quadruplet: batch index, x,y and feature index (x=y=0 for matrices)).
    /// Let h0_j be the recurrent input (previous output) data from previous time step (at time index t-1 of sequence).
    /// Let h_i be the proposed new output.
    /// Let h1_i be the output data produced at this time step.
    ///
    /// Let Wz_ij, Uz_ij, be the input gate weights for input and recurrent input data respectively
    /// Let bi_i be the bias for the input gate
    ///
    /// Let Wr_ij, Ur_ij be the recurrent gate weights for input and recurrent input data respectively
    /// Let br_i be the bias for the recurrent gate
    ///
    /// Let Wh_ij, Uh_ij, Vh_ij, be the output gate weights for input, recurrent gate and input gate respectively
    /// Let bh_i be the bias for the output gate
    ///
    /// Let gz(x), gr(x), gh(x) be the neuron activation function for the input, recurrent and output gates
    /// Let p > 0 be a scalar variable (typicall p >= 1.0) that defines the p-norm gating norm value.
    ///
    /// Then the output of the Gated Recurrent Unit layer is computed as follows:
    ///
    /// z_i = gz(  Wz_ij * x_j  +  Uz_ij * h0_j  +  bz_i  )
    /// r_i = gr(  Wr_ij * x_j  +  Ur_ij * h0_j  +  br_i  )
    /// c_i =      Uh_ij * (r_j h0_j)  +  Vh_ij * (z_j h0_j)
    /// h_i = gh(  Wh_ij * x_j  + c_i + bh_i  )
    ///
    /// h1_i = ( 1 - z_i ^ p)^(1/p) h_i + z_i h0_i
    ///
    /// The '*' stands for convolution (see
    /// MPSRNNImageInferenceLayer)or matrix-vector/matrix multiplication
    /// (see
    /// MPSRNNMatrixInferenceLayer).Summation is over index j (except for the batch index), but there is no summation over
    /// repeated index i - the output index.
    /// Note that for validity all intermediate images have to be of same size and all U and V matrices have to be square
    /// (ie. outputFeatureChannels == inputFeatureChannels in those). Also the bias terms are scalars wrt. spatial dimensions.
    /// The conventional GRU block is achieved by setting Vh = 0 (nil) and the so-called Minimal Gated Unit is achieved with Uh = 0.
    /// (The Minimal Gated Unit is detailed in: https://arxiv.org/abs/1603.09420 and there they call z_i the value of the forget gate).
    ///
    /// See also [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsgrudescriptor?language=objc)
    #[unsafe(super(MPSRNNDescriptor, NSObject))]
    #[derive(Debug, PartialEq, Eq, Hash)]
    pub struct MPSGRUDescriptor;
);

extern_conformance!(
    unsafe impl NSObjectProtocol for MPSGRUDescriptor {}
);

impl MPSGRUDescriptor {
    extern_methods!(
        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Wz_ij', bias 'bz_i' and neuron 'gz' from the GRU formula.
        /// If nil then assumed zero weights, bias and no neuron (identity mapping). Defaults to nil.
        #[unsafe(method(inputGateInputWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn inputGateInputWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`inputGateInputWeights`][Self::inputGateInputWeights].
        #[unsafe(method(setInputGateInputWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setInputGateInputWeights(
            &self,
            input_gate_input_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Uz_ij' from the GRU formula.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(inputGateRecurrentWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn inputGateRecurrentWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`inputGateRecurrentWeights`][Self::inputGateRecurrentWeights].
        #[unsafe(method(setInputGateRecurrentWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setInputGateRecurrentWeights(
            &self,
            input_gate_recurrent_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Wr_ij', bias 'br_i' and neuron 'gr' from the GRU formula.
        /// If nil then assumed zero weights, bias and no neuron (identity mapping).Defaults to nil.
        #[unsafe(method(recurrentGateInputWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn recurrentGateInputWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`recurrentGateInputWeights`][Self::recurrentGateInputWeights].
        #[unsafe(method(setRecurrentGateInputWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setRecurrentGateInputWeights(
            &self,
            recurrent_gate_input_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Ur_ij' from the GRU formula.
        /// If nil then assumed zero weights.Defaults to nil.
        #[unsafe(method(recurrentGateRecurrentWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn recurrentGateRecurrentWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`recurrentGateRecurrentWeights`][Self::recurrentGateRecurrentWeights].
        #[unsafe(method(setRecurrentGateRecurrentWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setRecurrentGateRecurrentWeights(
            &self,
            recurrent_gate_recurrent_weights: Option<
                &ProtocolObject<dyn MPSCNNConvolutionDataSource>,
            >,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Wh_ij', bias 'bh_i' and neuron 'gh' from the GRU formula.
        /// If nil then assumed zero weights, bias and no neuron (identity mapping).Defaults to nil.
        #[unsafe(method(outputGateInputWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn outputGateInputWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`outputGateInputWeights`][Self::outputGateInputWeights].
        #[unsafe(method(setOutputGateInputWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setOutputGateInputWeights(
            &self,
            output_gate_input_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Uh_ij' from the GRU formula.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(outputGateRecurrentWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn outputGateRecurrentWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`outputGateRecurrentWeights`][Self::outputGateRecurrentWeights].
        #[unsafe(method(setOutputGateRecurrentWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setOutputGateRecurrentWeights(
            &self,
            output_gate_recurrent_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Vh_ij' - can be used to implement the "Minimally Gated Unit".
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(outputGateInputGateWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn outputGateInputGateWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`outputGateInputGateWeights`][Self::outputGateInputGateWeights].
        #[unsafe(method(setOutputGateInputGateWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setOutputGateInputGateWeights(
            &self,
            output_gate_input_gate_weights: Option<
                &ProtocolObject<dyn MPSCNNConvolutionDataSource>,
            >,
        );

        /// The p-norm gating norm value as specified by the GRU formulae. Defaults to 1.0f.
        #[unsafe(method(gatePnormValue))]
        #[unsafe(method_family = none)]
        pub unsafe fn gatePnormValue(&self) -> c_float;

        /// Setter for [`gatePnormValue`][Self::gatePnormValue].
        #[unsafe(method(setGatePnormValue:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setGatePnormValue(&self, gate_pnorm_value: c_float);

        /// If YES then the GRU-block output formula is changed to:
        /// h1_i = ( 1 - z_i ^ p)^(1/p) h0_i + z_i h_i.
        /// Defaults to NO.
        #[unsafe(method(flipOutputGates))]
        #[unsafe(method_family = none)]
        pub unsafe fn flipOutputGates(&self) -> bool;

        /// Setter for [`flipOutputGates`][Self::flipOutputGates].
        #[unsafe(method(setFlipOutputGates:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setFlipOutputGates(&self, flip_output_gates: bool);

        /// Creates a GRU descriptor.
        ///
        /// Parameter `inputFeatureChannels`: The number of feature channels in the input image/matrix. Must be >= 1.
        ///
        /// Parameter `outputFeatureChannels`: The number of feature channels in the output image/matrix. Must be >= 1.
        ///
        /// Returns: A valid MPSGRUDescriptor object or nil, if failure.
        #[unsafe(method(createGRUDescriptorWithInputFeatureChannels:outputFeatureChannels:))]
        #[unsafe(method_family = none)]
        pub unsafe fn createGRUDescriptorWithInputFeatureChannels_outputFeatureChannels(
            input_feature_channels: NSUInteger,
            output_feature_channels: NSUInteger,
        ) -> Retained<Self>;
    );
}

/// Methods declared on superclass `NSObject`.
impl MPSGRUDescriptor {
    extern_methods!(
        #[unsafe(method(init))]
        #[unsafe(method_family = init)]
        pub unsafe fn init(this: Allocated<Self>) -> Retained<Self>;

        #[unsafe(method(new))]
        #[unsafe(method_family = new)]
        pub unsafe fn new() -> Retained<Self>;
    );
}

extern_class!(
    /// Dependencies: This depends on Metal.framework
    ///
    /// The MPSLSTMDescriptor specifies a LSTM block/layer descriptor.
    /// The RNN layer initialized with a MPSLSTMDescriptor transforms the input data (image or matrix),
    /// the memory cell data and previous output with a set of filters, each producing one feature map in
    /// the output data and memory cell, according to the LSTM formulae detailed below.
    /// The user may provide the LSTM unit a single input or a sequence of inputs.
    ///
    /// Description of operation:
    ///
    /// Let x_j be the input data (at time index t of sequence,
    /// j index containing quadruplet: batch index, x,y and feature index (x=y=0 for matrices)).
    /// Let h0_j be the recurrent input (previous output) data from previous time step (at time index t-1 of sequence).
    /// Let h1_i be the output data produced at this time step.
    /// Let c0_j be the previous memory cell data (at time index t-1 of sequence).
    /// Let c1_i be the new memory cell data (at time index t-1 of sequence).
    ///
    /// Let Wi_ij, Ui_ij, Vi_ij, be the input gate weights for input, recurrent input and memory cell (peephole) data respectively
    /// Let bi_i be the bias for the input gate
    ///
    /// Let Wf_ij, Uf_ij, Vf_ij, be the forget gate weights for input, recurrent input and memory cell data respectively
    /// Let bf_i be the bias for the forget gate
    ///
    /// Let Wo_ij, Uo_ij, Vo_ij, be the output gate weights for input, recurrent input and memory cell data respectively
    /// Let bo_i be the bias for the output gate
    ///
    /// Let Wc_ij, Uc_ij, Vc_ij, be the memory cell gate weights for input, recurrent input and memory cell data respectively
    /// Let bc_i be the bias for the memory cell gate
    ///
    /// Let gi(x), gf(x), go(x), gc(x) be neuron activation function for the input, forget, output gate and memory cell gate
    /// Let gh(x) be the activation function applied to result memory cell data
    ///
    /// Then the new memory cell data c1_j and output image h1_i are computed as follows:
    ///
    /// I_i = gi(  Wi_ij * x_j  +  Ui_ij * h0_j  +  Vi_ij * c0_j  + bi_i  )
    /// F_i = gf(  Wf_ij * x_j  +  Uf_ij * h0_j  +  Vf_ij * c0_j  + bf_i  )
    /// C_i = gc(  Wc_ij * x_j  +  Uc_ij * h0_j  +  Vc_ij * c0_j  + bc_i  )
    ///
    /// c1_i = F_i c0_i  +  I_i C_i
    ///
    /// O_i = go(  Wo_ij * x_j  +  Uo_ij * h0_j  +  Vo_ij * c1_j  + bo_i  )
    ///
    /// h1_i = O_i gh( c1_i )
    ///
    /// The '*' stands for convolution (see
    /// MPSRNNImageInferenceLayer)or matrix-vector/matrix multiplication
    /// (see
    /// MPSRNNMatrixInferenceLayer).Summation is over index j (except for the batch index), but there is no summation over
    /// repeated index i - the output index.
    /// Note that for validity all intermediate images have to be of same size and all U and V matrices have to be square
    /// (ie. outputFeatureChannels == inputFeatureChannels in those). Also the bias terms are scalars wrt. spatial dimensions.
    ///
    /// See also [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpslstmdescriptor?language=objc)
    #[unsafe(super(MPSRNNDescriptor, NSObject))]
    #[derive(Debug, PartialEq, Eq, Hash)]
    pub struct MPSLSTMDescriptor;
);

extern_conformance!(
    unsafe impl NSObjectProtocol for MPSLSTMDescriptor {}
);

impl MPSLSTMDescriptor {
    extern_methods!(
        /// If YES, then the 'peephole' weight matrices will be diagonal matrices represented as
        /// vectors of length the number of features in memory cells, that will be multiplied pointwise
        /// with the peephole matrix or image in order to achieve the diagonal (nonmixing) update.
        /// Defaults to NO.
        #[unsafe(method(memoryWeightsAreDiagonal))]
        #[unsafe(method_family = none)]
        pub unsafe fn memoryWeightsAreDiagonal(&self) -> bool;

        /// Setter for [`memoryWeightsAreDiagonal`][Self::memoryWeightsAreDiagonal].
        #[unsafe(method(setMemoryWeightsAreDiagonal:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setMemoryWeightsAreDiagonal(&self, memory_weights_are_diagonal: bool);

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Wi_ij', bias 'bi_i' and neuron 'gi' from the LSTM formula.
        /// If nil then assumed zero weights, bias and no neuron (identity mapping). Defaults to nil.
        #[unsafe(method(inputGateInputWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn inputGateInputWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`inputGateInputWeights`][Self::inputGateInputWeights].
        #[unsafe(method(setInputGateInputWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setInputGateInputWeights(
            &self,
            input_gate_input_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Ui_ij' from the LSTM formula.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(inputGateRecurrentWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn inputGateRecurrentWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`inputGateRecurrentWeights`][Self::inputGateRecurrentWeights].
        #[unsafe(method(setInputGateRecurrentWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setInputGateRecurrentWeights(
            &self,
            input_gate_recurrent_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Vi_ij' - the 'peephole' weights - from the LSTM formula.
        /// if YES == memoryWeightsAreDiagonal, then the number of weights used is the number of features
        /// in the memory cell image/matrix.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(inputGateMemoryWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn inputGateMemoryWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`inputGateMemoryWeights`][Self::inputGateMemoryWeights].
        #[unsafe(method(setInputGateMemoryWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setInputGateMemoryWeights(
            &self,
            input_gate_memory_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Wf_ij', bias 'bf_i' and neuron 'gf' from the LSTM formula.
        /// If nil then assumed zero weights, bias and no neuron (identity mapping).Defaults to nil.
        #[unsafe(method(forgetGateInputWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn forgetGateInputWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`forgetGateInputWeights`][Self::forgetGateInputWeights].
        #[unsafe(method(setForgetGateInputWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setForgetGateInputWeights(
            &self,
            forget_gate_input_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Uf_ij' from the LSTM formula.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(forgetGateRecurrentWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn forgetGateRecurrentWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`forgetGateRecurrentWeights`][Self::forgetGateRecurrentWeights].
        #[unsafe(method(setForgetGateRecurrentWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setForgetGateRecurrentWeights(
            &self,
            forget_gate_recurrent_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Vf_ij' - the 'peephole' weights - from the LSTM formula.
        /// if YES == memoryWeightsAreDiagonal, then the number of weights used is the number of features
        /// in the memory cell image/matrix.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(forgetGateMemoryWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn forgetGateMemoryWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`forgetGateMemoryWeights`][Self::forgetGateMemoryWeights].
        #[unsafe(method(setForgetGateMemoryWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setForgetGateMemoryWeights(
            &self,
            forget_gate_memory_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Wo_ij', bias 'bo_i' and neuron 'go' from the LSTM formula.
        /// If nil then assumed zero weights, bias and no neuron (identity mapping). Defaults to nil.
        #[unsafe(method(outputGateInputWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn outputGateInputWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`outputGateInputWeights`][Self::outputGateInputWeights].
        #[unsafe(method(setOutputGateInputWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setOutputGateInputWeights(
            &self,
            output_gate_input_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Uo_ij' from the LSTM formula.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(outputGateRecurrentWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn outputGateRecurrentWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`outputGateRecurrentWeights`][Self::outputGateRecurrentWeights].
        #[unsafe(method(setOutputGateRecurrentWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setOutputGateRecurrentWeights(
            &self,
            output_gate_recurrent_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Vo_ij' - the 'peephole' weights - from the LSTM.
        /// if YES == memoryWeightsAreDiagonal, then the number of weights used is the number of features
        /// in the memory cell image/matrix.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(outputGateMemoryWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn outputGateMemoryWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`outputGateMemoryWeights`][Self::outputGateMemoryWeights].
        #[unsafe(method(setOutputGateMemoryWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setOutputGateMemoryWeights(
            &self,
            output_gate_memory_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Wc_ij', bias 'bc_i' and neuron 'gc' from the LSTM formula.
        /// If nil then assumed zero weights, bias and no neuron (identity mapping). Defaults to nil.
        #[unsafe(method(cellGateInputWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn cellGateInputWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`cellGateInputWeights`][Self::cellGateInputWeights].
        #[unsafe(method(setCellGateInputWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setCellGateInputWeights(
            &self,
            cell_gate_input_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Uc_ij' from the LSTM formula.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(cellGateRecurrentWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn cellGateRecurrentWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`cellGateRecurrentWeights`][Self::cellGateRecurrentWeights].
        #[unsafe(method(setCellGateRecurrentWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setCellGateRecurrentWeights(
            &self,
            cell_gate_recurrent_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNConvolution")]
        /// Contains weights 'Vc_ij' - the 'peephole' weights - from the LSTM formula.
        /// if YES == memoryWeightsAreDiagonal, then the number of weights used is the number of features
        /// in the memory cell image/matrix.
        /// If nil then assumed zero weights. Defaults to nil.
        #[unsafe(method(cellGateMemoryWeights))]
        #[unsafe(method_family = none)]
        pub unsafe fn cellGateMemoryWeights(
            &self,
        ) -> Option<Retained<ProtocolObject<dyn MPSCNNConvolutionDataSource>>>;

        #[cfg(feature = "MPSCNNConvolution")]
        /// Setter for [`cellGateMemoryWeights`][Self::cellGateMemoryWeights].
        #[unsafe(method(setCellGateMemoryWeights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setCellGateMemoryWeights(
            &self,
            cell_gate_memory_weights: Option<&ProtocolObject<dyn MPSCNNConvolutionDataSource>>,
        );

        #[cfg(feature = "MPSCNNNeuronType")]
        /// Neuron type definition for 'gh', see
        /// MPSCNNNeuronType.Defaults to MPSCNNNeuronTypeTanH.
        #[unsafe(method(cellToOutputNeuronType))]
        #[unsafe(method_family = none)]
        pub unsafe fn cellToOutputNeuronType(&self) -> MPSCNNNeuronType;

        #[cfg(feature = "MPSCNNNeuronType")]
        /// Setter for [`cellToOutputNeuronType`][Self::cellToOutputNeuronType].
        #[unsafe(method(setCellToOutputNeuronType:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setCellToOutputNeuronType(
            &self,
            cell_to_output_neuron_type: MPSCNNNeuronType,
        );

        /// Neuron parameter A for 'gh'. Defaults to 1.0f.
        #[unsafe(method(cellToOutputNeuronParamA))]
        #[unsafe(method_family = none)]
        pub unsafe fn cellToOutputNeuronParamA(&self) -> c_float;

        /// Setter for [`cellToOutputNeuronParamA`][Self::cellToOutputNeuronParamA].
        #[unsafe(method(setCellToOutputNeuronParamA:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setCellToOutputNeuronParamA(&self, cell_to_output_neuron_param_a: c_float);

        /// Neuron parameter B for 'gh'. Defaults to 1.0f.
        #[unsafe(method(cellToOutputNeuronParamB))]
        #[unsafe(method_family = none)]
        pub unsafe fn cellToOutputNeuronParamB(&self) -> c_float;

        /// Setter for [`cellToOutputNeuronParamB`][Self::cellToOutputNeuronParamB].
        #[unsafe(method(setCellToOutputNeuronParamB:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setCellToOutputNeuronParamB(&self, cell_to_output_neuron_param_b: c_float);

        /// Neuron parameter C for 'gh'. Defaults to 1.0f.
        #[unsafe(method(cellToOutputNeuronParamC))]
        #[unsafe(method_family = none)]
        pub unsafe fn cellToOutputNeuronParamC(&self) -> c_float;

        /// Setter for [`cellToOutputNeuronParamC`][Self::cellToOutputNeuronParamC].
        #[unsafe(method(setCellToOutputNeuronParamC:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setCellToOutputNeuronParamC(&self, cell_to_output_neuron_param_c: c_float);

        /// Creates a LSTM descriptor.
        ///
        /// Parameter `inputFeatureChannels`: The number of feature channels in the input image/matrix. Must be >= 1.
        ///
        /// Parameter `outputFeatureChannels`: The number of feature channels in the output image/matrix. Must be >= 1.
        ///
        /// Returns: A valid MPSNNLSTMDescriptor object or nil, if failure.
        #[unsafe(method(createLSTMDescriptorWithInputFeatureChannels:outputFeatureChannels:))]
        #[unsafe(method_family = none)]
        pub unsafe fn createLSTMDescriptorWithInputFeatureChannels_outputFeatureChannels(
            input_feature_channels: NSUInteger,
            output_feature_channels: NSUInteger,
        ) -> Retained<Self>;
    );
}

/// Methods declared on superclass `NSObject`.
impl MPSLSTMDescriptor {
    extern_methods!(
        #[unsafe(method(init))]
        #[unsafe(method_family = init)]
        pub unsafe fn init(this: Allocated<Self>) -> Retained<Self>;

        #[unsafe(method(new))]
        #[unsafe(method_family = new)]
        pub unsafe fn new() -> Retained<Self>;
    );
}

extern_class!(
    /// Dependencies: This depends on Metal.framework
    ///
    /// This class holds all the data that is passed from one sequence iteration of the image-based RNN layer (stack) to the next.
    ///
    /// See also [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnnrecurrentimagestate?language=objc)
    #[unsafe(super(MPSState, NSObject))]
    #[derive(Debug, PartialEq, Eq, Hash)]
    #[cfg(all(feature = "MPSCore", feature = "MPSState"))]
    pub struct MPSRNNRecurrentImageState;
);

#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
extern_conformance!(
    unsafe impl NSObjectProtocol for MPSRNNRecurrentImageState {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
impl MPSRNNRecurrentImageState {
    extern_methods!(
        #[cfg(feature = "MPSImage")]
        /// Access the stored recurrent image data.
        ///
        /// Parameter `layerIndex`: Index of the layer whose to get - belongs to { 0, 1,...,
        ///
        /// See: numberOfLayers - 1 }
        ///
        /// Returns: For valid layerIndex the recurrent output image data, otherwise nil.
        #[unsafe(method(getRecurrentOutputImageForLayerIndex:))]
        #[unsafe(method_family = none)]
        pub unsafe fn getRecurrentOutputImageForLayerIndex(
            &self,
            layer_index: NSUInteger,
        ) -> Option<Retained<MPSImage>>;

        #[cfg(feature = "MPSImage")]
        /// Access the stored memory cell image data (if present).
        ///
        /// Parameter `layerIndex`: Index of the layer whose to get - belongs to { 0, 1,...,
        ///
        /// See: numberOfLayers - 1 }
        ///
        /// Returns: For valid layerIndex the memory cell image data, otherwise nil.
        #[unsafe(method(getMemoryCellImageForLayerIndex:))]
        #[unsafe(method_family = none)]
        pub unsafe fn getMemoryCellImageForLayerIndex(
            &self,
            layer_index: NSUInteger,
        ) -> Option<Retained<MPSImage>>;
    );
}

/// Methods declared on superclass `MPSState`.
#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
impl MPSRNNRecurrentImageState {
    extern_methods!(
        /// Create a MPSState holding a temporary MTLBuffer
        ///
        /// Parameter `cmdBuf`: The command buffer against which the temporary resource is allocated
        ///
        /// Parameter `bufferSize`: The size of the buffer in bytes
        #[unsafe(method(temporaryStateWithCommandBuffer:bufferSize:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer_bufferSize(
            cmd_buf: &ProtocolObject<dyn MTLCommandBuffer>,
            buffer_size: usize,
        ) -> Retained<Self>;

        /// Create a MPSState holding a temporary MTLTexture
        ///
        /// Parameter `cmdBuf`: The command buffer against which the temporary resource is allocated
        ///
        /// Parameter `descriptor`: A descriptor for the new temporary texture
        #[unsafe(method(temporaryStateWithCommandBuffer:textureDescriptor:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer_textureDescriptor(
            cmd_buf: &ProtocolObject<dyn MTLCommandBuffer>,
            descriptor: &MTLTextureDescriptor,
        ) -> Retained<Self>;

        /// Create a new autoreleased temporary state object without underlying resource
        ///
        /// Parameter `cmdBuf`: The command buffer with which the temporary resource is associated
        #[unsafe(method(temporaryStateWithCommandBuffer:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer(
            cmd_buf: &ProtocolObject<dyn MTLCommandBuffer>,
        ) -> Retained<Self>;

        #[unsafe(method(initWithDevice:bufferSize:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_bufferSize(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            buffer_size: usize,
        ) -> Retained<Self>;

        #[unsafe(method(initWithDevice:textureDescriptor:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_textureDescriptor(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            descriptor: &MTLTextureDescriptor,
        ) -> Retained<Self>;

        /// Create a MPSState with a non-temporary MTLResource
        ///
        /// Parameter `resource`: A MTLBuffer or MTLTexture. May be nil.
        ///
        /// # Safety
        ///
        /// - `resource` may need to be synchronized.
        /// - `resource` may be unretained, you must ensure it is kept alive while in use.
        #[unsafe(method(initWithResource:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithResource(
            this: Allocated<Self>,
            resource: Option<&ProtocolObject<dyn MTLResource>>,
        ) -> Retained<Self>;

        #[unsafe(method(init))]
        #[unsafe(method_family = init)]
        pub unsafe fn init(this: Allocated<Self>) -> Option<Retained<Self>>;

        /// Initialize a non-temporary state to hold a number of textures and buffers
        ///
        /// The allocation of each resource will be deferred  until it is needed.
        /// This occurs when -resource or -resourceAtIndex: is called.
        ///
        /// Parameter `resourceList`: The list of resources to create.
        #[unsafe(method(initWithDevice:resourceList:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_resourceList(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            resource_list: &MPSStateResourceList,
        ) -> Retained<Self>;

        /// Initialize a temporary state to hold a number of textures and buffers
        ///
        /// The textures occur first in sequence
        #[unsafe(method(temporaryStateWithCommandBuffer:resourceList:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer_resourceList(
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            resource_list: &MPSStateResourceList,
        ) -> Retained<Self>;

        /// Create a state object with a list of MTLResources
        ///
        /// Because MPS prefers deferred allocation of resources
        /// your application should use -initWithTextures:bufferSizes:bufferCount:
        /// whenever possible. This method is useful for cases when the
        /// MTLResources must be initialized by the CPU.
        ///
        /// # Safety
        ///
        /// - `resources` generic may need to be synchronized.
        /// - `resources` generic may be unretained, you must ensure it is kept alive while in use.
        #[unsafe(method(initWithResources:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithResources(
            this: Allocated<Self>,
            resources: Option<&NSArray<ProtocolObject<dyn MTLResource>>>,
        ) -> Retained<Self>;
    );
}

/// Methods declared on superclass `NSObject`.
#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
impl MPSRNNRecurrentImageState {
    extern_methods!(
        #[unsafe(method(new))]
        #[unsafe(method_family = new)]
        pub unsafe fn new() -> Retained<Self>;
    );
}

extern_class!(
    /// Dependencies: This depends on Metal.framework
    ///
    /// The MPSRNNImageInferenceLayer specifies a recurrent neural network layer for inference on MPSImages.
    /// Currently two types of recurrent layers are supported: ones that operate with convolutions on
    /// images:
    /// MPSRNNImageInferenceLayerand one that operates on matrices:
    /// MPSRNNMatrixInferenceLayer.The former can be often used to implement the latter by using 1x1-images, but due to
    /// image size restrictions and performance, it is advisable to use
    /// MPSRNNMatrixInferenceLayerfor
    /// linear recurrent layers.
    /// A MPSRNNImageInferenceLayer is initialized using a
    /// MPSRNNLayerDescriptor,which further specifies the
    /// recurrent network layer, or an array of
    /// MPSRNNLayerDescriptors,which specifies a stack
    /// of recurrent layers, that can operate in parallel a subset of the inputs in a sequence of inputs and
    /// recurrent outputs. Note that currently stacks with bidirectionally traversing encode functions do not support starting
    /// from a previous set of recurrent states, but this can be achieved quite easily by defining two separate
    /// unidirectional stacks of layers, and running the same input sequence on them separately (one forwards and one backwards)
    /// and ultimately combining the two result sequences as desired with auxiliary functions.
    ///
    /// See also [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnnimageinferencelayer?language=objc)
    #[unsafe(super(MPSCNNKernel, MPSKernel, NSObject))]
    #[derive(Debug, PartialEq, Eq, Hash)]
    #[cfg(all(feature = "MPSCNNKernel", feature = "MPSCore", feature = "MPSKernel"))]
    pub struct MPSRNNImageInferenceLayer;
);

#[cfg(all(feature = "MPSCNNKernel", feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSCoding for MPSRNNImageInferenceLayer {}
);

#[cfg(all(feature = "MPSCNNKernel", feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSCopying for MPSRNNImageInferenceLayer {}
);

#[cfg(all(feature = "MPSCNNKernel", feature = "MPSCore", feature = "MPSKernel"))]
unsafe impl CopyingHelper for MPSRNNImageInferenceLayer {
    type Result = Self;
}

#[cfg(all(feature = "MPSCNNKernel", feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSObjectProtocol for MPSRNNImageInferenceLayer {}
);

#[cfg(all(feature = "MPSCNNKernel", feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSSecureCoding for MPSRNNImageInferenceLayer {}
);

#[cfg(all(feature = "MPSCNNKernel", feature = "MPSCore", feature = "MPSKernel"))]
impl MPSRNNImageInferenceLayer {
    extern_methods!(
        /// The number of feature channels per pixel in the input image.
        #[unsafe(method(inputFeatureChannels))]
        #[unsafe(method_family = none)]
        pub unsafe fn inputFeatureChannels(&self) -> NSUInteger;

        /// The number of feature channels per pixel in the output image.
        #[unsafe(method(outputFeatureChannels))]
        #[unsafe(method_family = none)]
        pub unsafe fn outputFeatureChannels(&self) -> NSUInteger;

        /// Number of layers in the filter-stack. This will be one when using initWithDevice:rnnDescriptor to initialize
        /// this filter and the number of entries in the array 'rnnDescriptors' when initializing this filter with
        /// initWithDevice:rnnDescriptors.
        #[unsafe(method(numberOfLayers))]
        #[unsafe(method_family = none)]
        pub unsafe fn numberOfLayers(&self) -> NSUInteger;

        /// How output states from
        /// encodeSequenceToCommandBufferare constructed.
        /// Defaults to NO. For reference
        ///
        /// See: MPSState.
        #[unsafe(method(recurrentOutputIsTemporary))]
        #[unsafe(method_family = none)]
        pub unsafe fn recurrentOutputIsTemporary(&self) -> bool;

        /// Setter for [`recurrentOutputIsTemporary`][Self::recurrentOutputIsTemporary].
        #[unsafe(method(setRecurrentOutputIsTemporary:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setRecurrentOutputIsTemporary(&self, recurrent_output_is_temporary: bool);

        /// If YES then calls to
        /// encodeSequenceToCommandBufferreturn every recurrent state
        /// in the array: recurrentOutputStates.
        /// Defaults to NO.
        #[unsafe(method(storeAllIntermediateStates))]
        #[unsafe(method_family = none)]
        pub unsafe fn storeAllIntermediateStates(&self) -> bool;

        /// Setter for [`storeAllIntermediateStates`][Self::storeAllIntermediateStates].
        #[unsafe(method(setStoreAllIntermediateStates:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setStoreAllIntermediateStates(&self, store_all_intermediate_states: bool);

        /// Defines how to combine the output-results, when encoding bidirectional layers using
        /// encodeBidirectionalSequenceToCommandBuffer.Defaults to
        /// MPSRNNBidirectionalCombineModeNone.
        #[unsafe(method(bidirectionalCombineMode))]
        #[unsafe(method_family = none)]
        pub unsafe fn bidirectionalCombineMode(&self) -> MPSRNNBidirectionalCombineMode;

        /// Setter for [`bidirectionalCombineMode`][Self::bidirectionalCombineMode].
        #[unsafe(method(setBidirectionalCombineMode:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setBidirectionalCombineMode(
            &self,
            bidirectional_combine_mode: MPSRNNBidirectionalCombineMode,
        );

        /// Initializes a convolutional RNN kernel
        ///
        /// Parameter `device`: The MTLDevice on which this MPSRNNImageLayer filter will be used
        ///
        /// Parameter `rnnDescriptor`: The descriptor that defines the RNN layer
        ///
        /// Returns: A valid MPSRNNImageInferenceLayer object or nil, if failure.
        #[unsafe(method(initWithDevice:rnnDescriptor:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_rnnDescriptor(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            rnn_descriptor: &MPSRNNDescriptor,
        ) -> Retained<Self>;

        /// Initializes a kernel that implements a stack of convolutional RNN layers
        ///
        /// Parameter `device`: The MTLDevice on which this MPSRNNImageLayer filter will be used
        ///
        /// Parameter `rnnDescriptors`: An array of RNN descriptors that defines a stack of RNN layers, starting at index zero.
        /// The number of layers in stack is the number of entries in the array.
        /// All entries in the array must be valid MPSRNNDescriptors.
        ///
        /// Returns: A valid MPSRNNImageInferenceLayer object or nil, if failure.
        #[unsafe(method(initWithDevice:rnnDescriptors:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_rnnDescriptors(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            rnn_descriptors: &NSArray<MPSRNNDescriptor>,
        ) -> Retained<Self>;

        #[unsafe(method(initWithDevice:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
        ) -> Retained<Self>;

        #[cfg(all(feature = "MPSImage", feature = "MPSState"))]
        /// Encode an MPSRNNImageInferenceLayer kernel (stack) for a sequence of inputs into a command buffer.
        /// Note that when encoding using this function the
        ///
        /// See: layerSequenceDirection is ignored and the layer stack operates as
        /// if all layers were forward feeding layers. In order to run bidirectional sequences
        /// use
        /// encodeBidirectionalSequenceToCommandBuffer:sourceSequence:or alternatively run two layer stacks and combine
        /// results at the end using utility functions.
        ///
        /// Parameter `commandBuffer`: A valid MTLCommandBuffer to receive the encoded filter
        ///
        /// Parameter `sourceImages`: An array of valid MPSImage objects containing the sequence of source images.
        ///
        /// Parameter `destinationImages`: An array valid MPSImages to be overwritten by result image sequence. destinationImages may not alias sourceImages.
        ///
        /// Parameter `recurrentInputState`: An optional state containing the output images and memory cells (for LSTMs)
        /// of the layer obtained from the previous input images in a sequence of inputs.
        /// Has to be the output of a previous call to this function or nil (assumed zero).
        /// Note: can be one of the states returned in
        /// recurrentOutputStates.
        /// Parameter `recurrentOutputStates`: An optional array that will contain the recurrent output states. If nil then
        /// the recurrent output state is discarded.
        /// If
        /// storeAllIntermediateStatesis YES, then all intermediate states of the sequence
        /// are returned in the array, the first one corresponding to the first input in the sequence,
        /// otherwise only the last recurrent output state is returned.
        /// If recurrentOutputIsTemporary is YES and then all returned recurrent states
        /// will be temporary.
        ///
        /// See: MPSState:isTemporary.
        /// Example: In order to get a new state one can do the following:
        ///
        /// ```text
        ///                                                       MPSRNNRecurrentImageState* recurrent0 = nil;
        ///                                                       [filter encodeToCommandBuffer: cmdBuf
        ///                                                                         sourceImage: source0
        ///                                                                    destinationImage: destination0
        ///                                                                 recurrentInputState: nil
        ///                                                                recurrentOutputState: &recurrent0];
        /// ```
        ///
        /// Then use it for the next input in sequence:
        ///
        /// ```text
        ///                                                       [filter encodeToCommandBuffer: cmdBuf
        ///                                                                         sourceImage: source1
        ///                                                                    destinationImage: destination1
        ///                                                                 recurrentInputState: recurrent0
        ///                                                                recurrentOutputState: &recurrent0];
        /// ```
        ///
        /// And discard recurrent output of the third input:
        ///
        /// ```text
        ///                                                       [filter encodeToCommandBuffer: cmdBuf
        ///                                                                         sourceImage: source2
        ///                                                                    destinationImage: destination2
        ///                                                                 recurrentInputState: recurrent0
        ///                                                                recurrentOutputState: nil];
        /// ```
        #[unsafe(method(encodeSequenceToCommandBuffer:sourceImages:destinationImages:recurrentInputState:recurrentOutputStates:))]
        #[unsafe(method_family = none)]
        pub unsafe fn encodeSequenceToCommandBuffer_sourceImages_destinationImages_recurrentInputState_recurrentOutputStates(
            &self,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            source_images: &NSArray<MPSImage>,
            destination_images: &NSArray<MPSImage>,
            recurrent_input_state: Option<&MPSRNNRecurrentImageState>,
            recurrent_output_states: Option<&NSMutableArray<MPSRNNRecurrentImageState>>,
        );

        #[cfg(feature = "MPSImage")]
        /// Encode an MPSRNNImageInferenceLayer kernel stack for an input image sequences into a command buffer bidirectionally.
        /// The operation proceeds as follows: The first source image x0 is passed through all forward traversing layers in the stack,
        /// ie. those that were initialized with MPSRNNSequenceDirectionForward, recurrent input is assumed zero.
        /// This produces forward output yf0 and recurrent states hf00, hf01, hf02, ... hf0n, one for each forward layer.
        /// Then x1 is passed to forward layers together with recurrent state hf00, hf01, ..., hf0n, which produces yf1, and hf10,...
        /// This procedure is iterated until the last image in the input sequence x_(N-1), which produces forward output yf(N-1).
        /// The backwards layers iterate the same sequence backwards, starting from input x_(N-1) (recurrent state zero),
        /// that produces yb(N-1) and recurrent output hb(N-1)0, hf(N-1)1, ... hb(N-1)m, one for each backwards traversing layer.
        /// Then the backwards layers handle input x_(N-2) using recurrent state hb(N-1)0, ..., et cetera, until the
        /// first image of the sequence is computed, producing output yb0. The result of the operation is either pair of sequences
        /// ({yf0, yf1, ... , yf(N-1)},  {yb0, yb1, ... , yb(N-1)}) or a combined sequence, {(yf0 + yb0), ... , (yf(N-1) + yb(N-1)) },
        /// where '+' stands either for sum, or concatenation along feature channels, as specified by
        /// bidirectionalCombineMode.
        ///
        /// Parameter `commandBuffer`: A valid MTLCommandBuffer to receive the encoded filter
        ///
        /// Parameter `sourceSequence`: An array of valid MPSImage objects containing the source image sequence (x0, x1, ... x_n-1).
        ///
        /// Parameter `destinationForwardImages`: An array of valid MPSImages to be overwritten by result from forward input images. If bidirectionalCombineMode
        /// is either MPSRNNBidirectionalCombineModeAdd or MPSRNNBidirectionalCombineModeConcatenate, then will
        /// contain the combined results. destinationForwardImage may not alias with any of the source images.
        ///
        /// Parameter `destinationBackwardImages`: If bidirectionalCombineMode is MPSRNNBidirectionalCombineModeNone, then must be a valid MPSImage
        /// that will be  overwritten by result from backward input image. Otherwise this parameter is ignored
        /// and can be nil. destinationBackwardImages may not alias to any of the source images.
        #[unsafe(method(encodeBidirectionalSequenceToCommandBuffer:sourceSequence:destinationForwardImages:destinationBackwardImages:))]
        #[unsafe(method_family = none)]
        pub unsafe fn encodeBidirectionalSequenceToCommandBuffer_sourceSequence_destinationForwardImages_destinationBackwardImages(
            &self,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            source_sequence: &NSArray<MPSImage>,
            destination_forward_images: &NSArray<MPSImage>,
            destination_backward_images: Option<&NSArray<MPSImage>>,
        );

        /// NSSecureCoding compatability
        ///
        /// See
        /// MPSKernel#initWithCoder.
        /// Parameter `aDecoder`: The NSCoder subclass with your serialized MPSRNNImageInferenceLayer
        ///
        /// Parameter `device`: The MTLDevice on which to make the MPSRNNImageInferenceLayer
        ///
        /// Returns: A new MPSRNNImageInferenceLayer object, or nil if failure.
        ///
        /// # Safety
        ///
        /// `a_decoder` possibly has further requirements.
        #[unsafe(method(initWithCoder:device:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithCoder_device(
            this: Allocated<Self>,
            a_decoder: &NSCoder,
            device: &ProtocolObject<dyn MTLDevice>,
        ) -> Option<Retained<Self>>;

        /// Make a copy of this kernel for a new device -
        ///
        /// See: MPSKernel
        ///
        /// Parameter `zone`: The NSZone in which to allocate the object
        ///
        /// Parameter `device`: The device for the new MPSKernel. If nil, then use
        /// self.device.
        ///
        /// Returns: a pointer to a copy of this MPSKernel. This will fail, returning
        /// nil if the device is not supported. Devices must be
        /// MTLFeatureSet_iOS_GPUFamily2_v1 or later.
        ///
        /// # Safety
        ///
        /// `zone` must be a valid pointer or null.
        #[unsafe(method(copyWithZone:device:))]
        #[unsafe(method_family = copy)]
        pub unsafe fn copyWithZone_device(
            &self,
            zone: *mut NSZone,
            device: Option<&ProtocolObject<dyn MTLDevice>>,
        ) -> Retained<Self>;
    );
}

/// Methods declared on superclass `MPSKernel`.
#[cfg(all(feature = "MPSCNNKernel", feature = "MPSCore", feature = "MPSKernel"))]
impl MPSRNNImageInferenceLayer {
    extern_methods!(
        /// Called by NSCoder to decode MPSKernels
        ///
        /// This isn't the right interface to decode a MPSKernel, but
        /// it is the one that NSCoder uses. To enable your NSCoder
        /// (e.g. NSKeyedUnarchiver) to set which device to use
        /// extend the object to adopt the MPSDeviceProvider
        /// protocol. Otherwise, the Metal system default device
        /// will be used.
        ///
        /// # Safety
        ///
        /// `a_decoder` possibly has further requirements.
        #[unsafe(method(initWithCoder:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithCoder(
            this: Allocated<Self>,
            a_decoder: &NSCoder,
        ) -> Option<Retained<Self>>;
    );
}

/// Methods declared on superclass `NSObject`.
#[cfg(all(feature = "MPSCNNKernel", feature = "MPSCore", feature = "MPSKernel"))]
impl MPSRNNImageInferenceLayer {
    extern_methods!(
        #[unsafe(method(init))]
        #[unsafe(method_family = init)]
        pub unsafe fn init(this: Allocated<Self>) -> Retained<Self>;

        #[unsafe(method(new))]
        #[unsafe(method_family = new)]
        pub unsafe fn new() -> Retained<Self>;
    );
}

extern_class!(
    /// Dependencies: This depends on Metal.framework
    ///
    /// This class holds all the data that is passed from one sequence iteration of the matrix-based RNN layer to the next.
    ///
    /// See also [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnnrecurrentmatrixstate?language=objc)
    #[unsafe(super(MPSState, NSObject))]
    #[derive(Debug, PartialEq, Eq, Hash)]
    #[cfg(all(feature = "MPSCore", feature = "MPSState"))]
    pub struct MPSRNNRecurrentMatrixState;
);

#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
extern_conformance!(
    unsafe impl NSObjectProtocol for MPSRNNRecurrentMatrixState {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
impl MPSRNNRecurrentMatrixState {
    extern_methods!(
        #[cfg(feature = "MPSMatrix")]
        /// Access the stored recurrent matrix data.
        ///
        /// Parameter `layerIndex`: Index of the layer whose to get - belongs to { 0, 1,...,
        ///
        /// See: numberOfLayers - 1 }
        ///
        /// Returns: For valid layerIndex the recurrent output matrix data, otherwise nil.
        #[unsafe(method(getRecurrentOutputMatrixForLayerIndex:))]
        #[unsafe(method_family = none)]
        pub unsafe fn getRecurrentOutputMatrixForLayerIndex(
            &self,
            layer_index: NSUInteger,
        ) -> Option<Retained<MPSMatrix>>;

        #[cfg(feature = "MPSMatrix")]
        /// Access the stored memory cell matrix data (if present).
        ///
        /// Parameter `layerIndex`: Index of the layer whose to get - belongs to { 0, 1,...,
        ///
        /// See: numberOfLayers - 1 }
        ///
        /// Returns: For valid layerIndex the memory cell image matrix, otherwise nil.
        #[unsafe(method(getMemoryCellMatrixForLayerIndex:))]
        #[unsafe(method_family = none)]
        pub unsafe fn getMemoryCellMatrixForLayerIndex(
            &self,
            layer_index: NSUInteger,
        ) -> Option<Retained<MPSMatrix>>;
    );
}

/// Methods declared on superclass `MPSState`.
#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
impl MPSRNNRecurrentMatrixState {
    extern_methods!(
        /// Create a MPSState holding a temporary MTLBuffer
        ///
        /// Parameter `cmdBuf`: The command buffer against which the temporary resource is allocated
        ///
        /// Parameter `bufferSize`: The size of the buffer in bytes
        #[unsafe(method(temporaryStateWithCommandBuffer:bufferSize:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer_bufferSize(
            cmd_buf: &ProtocolObject<dyn MTLCommandBuffer>,
            buffer_size: usize,
        ) -> Retained<Self>;

        /// Create a MPSState holding a temporary MTLTexture
        ///
        /// Parameter `cmdBuf`: The command buffer against which the temporary resource is allocated
        ///
        /// Parameter `descriptor`: A descriptor for the new temporary texture
        #[unsafe(method(temporaryStateWithCommandBuffer:textureDescriptor:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer_textureDescriptor(
            cmd_buf: &ProtocolObject<dyn MTLCommandBuffer>,
            descriptor: &MTLTextureDescriptor,
        ) -> Retained<Self>;

        /// Create a new autoreleased temporary state object without underlying resource
        ///
        /// Parameter `cmdBuf`: The command buffer with which the temporary resource is associated
        #[unsafe(method(temporaryStateWithCommandBuffer:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer(
            cmd_buf: &ProtocolObject<dyn MTLCommandBuffer>,
        ) -> Retained<Self>;

        #[unsafe(method(initWithDevice:bufferSize:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_bufferSize(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            buffer_size: usize,
        ) -> Retained<Self>;

        #[unsafe(method(initWithDevice:textureDescriptor:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_textureDescriptor(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            descriptor: &MTLTextureDescriptor,
        ) -> Retained<Self>;

        /// Create a MPSState with a non-temporary MTLResource
        ///
        /// Parameter `resource`: A MTLBuffer or MTLTexture. May be nil.
        ///
        /// # Safety
        ///
        /// - `resource` may need to be synchronized.
        /// - `resource` may be unretained, you must ensure it is kept alive while in use.
        #[unsafe(method(initWithResource:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithResource(
            this: Allocated<Self>,
            resource: Option<&ProtocolObject<dyn MTLResource>>,
        ) -> Retained<Self>;

        #[unsafe(method(init))]
        #[unsafe(method_family = init)]
        pub unsafe fn init(this: Allocated<Self>) -> Option<Retained<Self>>;

        /// Initialize a non-temporary state to hold a number of textures and buffers
        ///
        /// The allocation of each resource will be deferred  until it is needed.
        /// This occurs when -resource or -resourceAtIndex: is called.
        ///
        /// Parameter `resourceList`: The list of resources to create.
        #[unsafe(method(initWithDevice:resourceList:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_resourceList(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            resource_list: &MPSStateResourceList,
        ) -> Retained<Self>;

        /// Initialize a temporary state to hold a number of textures and buffers
        ///
        /// The textures occur first in sequence
        #[unsafe(method(temporaryStateWithCommandBuffer:resourceList:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer_resourceList(
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            resource_list: &MPSStateResourceList,
        ) -> Retained<Self>;

        /// Create a state object with a list of MTLResources
        ///
        /// Because MPS prefers deferred allocation of resources
        /// your application should use -initWithTextures:bufferSizes:bufferCount:
        /// whenever possible. This method is useful for cases when the
        /// MTLResources must be initialized by the CPU.
        ///
        /// # Safety
        ///
        /// - `resources` generic may need to be synchronized.
        /// - `resources` generic may be unretained, you must ensure it is kept alive while in use.
        #[unsafe(method(initWithResources:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithResources(
            this: Allocated<Self>,
            resources: Option<&NSArray<ProtocolObject<dyn MTLResource>>>,
        ) -> Retained<Self>;
    );
}

/// Methods declared on superclass `NSObject`.
#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
impl MPSRNNRecurrentMatrixState {
    extern_methods!(
        #[unsafe(method(new))]
        #[unsafe(method_family = new)]
        pub unsafe fn new() -> Retained<Self>;
    );
}

extern_class!(
    /// Dependencies: This depends on Metal.framework
    ///
    /// The MPSRNNMatrixInferenceLayer specifies a recurrent neural network layer for inference on MPSMatrices.
    /// Currently two types of recurrent layers are supported: ones that operate with convolutions on
    /// images:
    /// MPSRNNImageInferenceLayerand one that operates on matrices:
    /// MPSRNNMatrixInferenceLayer.The former can be often used to implement the latter by using 1x1-matrices, but due to
    /// image size restrictions and performance, it is advisable to use
    /// MPSRNNMatrixInferenceLayerfor
    /// linear recurrent layers.
    /// A MPSRNNMatrixInferenceLayer is initialized using a
    /// MPSRNNLayerDescriptor,which further specifies the
    /// recurrent network layer, or an array of
    /// MPSRNNLayerDescriptors,which specifies a stack
    /// of recurrent layers, that can operate in parallel a subset of the inputs in a sequence of inputs and
    /// recurrent outputs. Note that currently stacks with bidirectionally traversing encode functions do not support starting
    /// from a previous set of recurrent states, but this can be achieved quite easily by defining two separate
    /// unidirectional stacks of layers, and running the same input sequence on them separately (one forwards and one backwards)
    /// and ultimately combining the two result sequences as desired with auxiliary functions.
    /// The input and output vectors in encode calls are stored as rows of the input and output matrices and
    /// MPSRNNMatrixInferenceLayer supports matrices with decreasing number of rows: The row-indices identify the different
    /// sequences that may be of different lengths - for example if we have three sequences:
    /// ( x1, x2, x3 ), ( y1, y2, y3, y4 ) and ( z1, z2 )
    /// of vectors xi, yi and zi, then these can be inserted together as a batch to the sequence encoding kernel by
    /// using the matrices:
    ///
    /// ```text
    ///                            ( y1 )        ( y2 )        ( y3 )        ( y4 )
    ///                       m1 = ( x1 ),  m2 = ( x2 ),  m3 = ( x3 ),  m4 =
    ///                            ( z1 )        ( z2 )
    /// ```
    ///
    /// If a recurrent output state is requested then it will contain the state corresponding to last inputs to each
    /// sequence and if all the intermediate states are requested (see storeAllIntermediateStates),
    /// then the shorter sequences will be propagated by copying the state of the previous output if the
    /// input vector is not present in the sequence - in the example above the output states would be:
    ///
    /// ```text
    ///                            ( s_y1 )        ( s_y2 )        ( s_y3 )        ( s_y4 )
    ///                       s1 = ( s_x1 ),  s2 = ( s_x2 ),  s3 = ( s_x3 ),  s4 = ( s_x3 )
    ///                            ( s_z1 )        ( s_z2 )        ( s_z2 )        ( s_z2 )
    /// ```
    ///
    /// The mathematical operation described in the linear transformations of
    /// MPSRNNSingleGateDescriptorMPSLSTMDescriptorand
    /// MPSGRUDescriptorare y^T = W x^T
    /// <
    /// => y = x W^T, where x is the matrix containing
    /// the input vectors as rows, y is the matrix containing the output vectors as rows and W is the weight matrix.
    ///
    /// See also [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnnmatrixinferencelayer?language=objc)
    #[unsafe(super(MPSKernel, NSObject))]
    #[derive(Debug, PartialEq, Eq, Hash)]
    #[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
    pub struct MPSRNNMatrixInferenceLayer;
);

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSCoding for MPSRNNMatrixInferenceLayer {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSCopying for MPSRNNMatrixInferenceLayer {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
unsafe impl CopyingHelper for MPSRNNMatrixInferenceLayer {
    type Result = Self;
}

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSObjectProtocol for MPSRNNMatrixInferenceLayer {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSSecureCoding for MPSRNNMatrixInferenceLayer {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
impl MPSRNNMatrixInferenceLayer {
    extern_methods!(
        /// The number of feature channels input vector/matrix.
        #[unsafe(method(inputFeatureChannels))]
        #[unsafe(method_family = none)]
        pub unsafe fn inputFeatureChannels(&self) -> NSUInteger;

        /// The number of feature channels in the output vector/matrix.
        #[unsafe(method(outputFeatureChannels))]
        #[unsafe(method_family = none)]
        pub unsafe fn outputFeatureChannels(&self) -> NSUInteger;

        /// Number of layers in the filter-stack. This will be one when using initWithDevice:rnnDescriptor to initialize
        /// this filter and the number of entries in the array 'rnnDescriptors' when initializing this filter with
        /// initWithDevice:rnnDescriptors.
        #[unsafe(method(numberOfLayers))]
        #[unsafe(method_family = none)]
        pub unsafe fn numberOfLayers(&self) -> NSUInteger;

        /// How output states from
        /// encodeSequenceToCommandBufferare constructed.
        /// Defaults to NO. For reference
        ///
        /// See: MPSState.
        #[unsafe(method(recurrentOutputIsTemporary))]
        #[unsafe(method_family = none)]
        pub unsafe fn recurrentOutputIsTemporary(&self) -> bool;

        /// Setter for [`recurrentOutputIsTemporary`][Self::recurrentOutputIsTemporary].
        #[unsafe(method(setRecurrentOutputIsTemporary:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setRecurrentOutputIsTemporary(&self, recurrent_output_is_temporary: bool);

        /// If YES then calls to
        /// encodeSequenceToCommandBufferreturn every recurrent state
        /// in the array: recurrentOutputStates.
        /// Defaults to NO.
        #[unsafe(method(storeAllIntermediateStates))]
        #[unsafe(method_family = none)]
        pub unsafe fn storeAllIntermediateStates(&self) -> bool;

        /// Setter for [`storeAllIntermediateStates`][Self::storeAllIntermediateStates].
        #[unsafe(method(setStoreAllIntermediateStates:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setStoreAllIntermediateStates(&self, store_all_intermediate_states: bool);

        /// Defines how to combine the output-results, when encoding bidirectional layers using
        /// encodeBidirectionalSequenceToCommandBuffer.Defaults to
        /// MPSRNNBidirectionalCombineModeNone.
        #[unsafe(method(bidirectionalCombineMode))]
        #[unsafe(method_family = none)]
        pub unsafe fn bidirectionalCombineMode(&self) -> MPSRNNBidirectionalCombineMode;

        /// Setter for [`bidirectionalCombineMode`][Self::bidirectionalCombineMode].
        #[unsafe(method(setBidirectionalCombineMode:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setBidirectionalCombineMode(
            &self,
            bidirectional_combine_mode: MPSRNNBidirectionalCombineMode,
        );

        /// Initializes a linear (fully connected) RNN kernel
        ///
        /// Parameter `device`: The MTLDevice on which this MPSRNNMatrixLayer filter will be used
        ///
        /// Parameter `rnnDescriptor`: The descriptor that defines the RNN layer
        ///
        /// Returns: A valid MPSRNNMatrixInferenceLayer object or nil, if failure.
        #[unsafe(method(initWithDevice:rnnDescriptor:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_rnnDescriptor(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            rnn_descriptor: &MPSRNNDescriptor,
        ) -> Retained<Self>;

        /// Initializes a kernel that implements a stack of linear (fully connected) RNN layers
        ///
        /// Parameter `device`: The MTLDevice on which this MPSRNNMatrixLayer filter will be used
        ///
        /// Parameter `rnnDescriptors`: An array of RNN descriptors that defines a stack of RNN layers, starting at index zero.
        /// The number of layers in stack is the number of entries in the array.
        /// All entries in the array must be valid MPSRNNDescriptors.
        ///
        /// Returns: A valid MPSRNNMatrixInferenceLayer object or nil, if failure.
        #[unsafe(method(initWithDevice:rnnDescriptors:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_rnnDescriptors(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            rnn_descriptors: &NSArray<MPSRNNDescriptor>,
        ) -> Retained<Self>;

        #[unsafe(method(initWithDevice:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
        ) -> Retained<Self>;

        #[cfg(all(feature = "MPSMatrix", feature = "MPSState"))]
        /// Encode an MPSRNNMatrixInferenceLayer kernel (stack) for a sequence of inputs into a command buffer.
        /// Note that when encoding using this function the
        ///
        /// See: layerSequenceDirection is ignored and the layer stack operates as
        /// if all layers were forward feeding layers. In order to run bidirectional sequences
        /// use
        /// encodeBidirectionalSequenceToCommandBuffer:sourceSequence:or alternatively run two layer stacks and combine
        /// results at the end using utility functions.
        ///
        /// Parameter `commandBuffer`: A valid MTLCommandBuffer to receive the encoded filter
        ///
        /// Parameter `sourceMatrices`: An array of valid MPSMatrix objects containing the sequence of source matrices.
        ///
        /// Parameter `sourceOffsets`: An array of byte-offsets into the sourceMatrices, if nil zeros are assumed and
        /// if not nil must contain offset for every matrix in sourceMatrices.
        ///
        /// Parameter `destinationMatrices`: An array valid MPSMatrices to be overwritten by result matrix sequence.
        /// destinationMatrices may not alias sourceMatrices.
        ///
        /// Parameter `destinationOffsets`: An array of byte-offsets into the destinationMatrices, if nil zeros are assumed and
        /// if not nil must contain offset for every matrix in destinationMatrices.
        ///
        /// Parameter `recurrentInputState`: An optional state containing the output matrices and memory cells (for LSTMs)
        /// of the layer obtained from the previous input matrices in a sequence of inputs.
        /// Has to be the output of a previous call to this function or nil (assumed zero).
        /// Note: can be one of the states returned in
        /// intermediateRecurrentStates.
        /// Parameter `recurrentOutputStates`: An optional array that will contain the recurrent output states. If nil then
        /// the recurrent output state is discarded.
        /// If
        /// storeAllIntermediateStatesis YES, then all intermediate states of the sequence
        /// are returned in the array, the first one corresponding to the first input in the sequence,
        /// otherwise only the last recurrent output state is returned.
        /// If recurrentOutputIsTemporary is YES and then all returned recurrent states
        /// will be temporary.
        ///
        /// See: MPSState:isTemporary.
        /// Example: In order to get a new state one can do the following:
        ///
        /// ```text
        ///                                                       MPSRNNRecurrentMatrixState* recurrent0 = nil;
        ///                                                       [filter encodeToCommandBuffer: cmdBuf
        ///                                                                        sourceMatrix: source0
        ///                                                                   destinationMatrix: destination0
        ///                                                                 recurrentInputState: nil
        ///                                                                recurrentOutputState: &recurrent0];
        /// ```
        ///
        /// Then use it for the next input in sequence:
        ///
        /// ```text
        ///                                                       [filter encodeToCommandBuffer: cmdBuf
        ///                                                                        sourceMatrix: source1
        ///                                                                   destinationMatrix: destination1
        ///                                                                 recurrentInputState: recurrent0
        ///                                                                recurrentOutputState: &recurrent0];
        /// ```
        ///
        /// And discard recurrent output of the third input:
        ///
        /// ```text
        ///                                                       [filter encodeToCommandBuffer: cmdBuf
        ///                                                                        sourceMatrix: source2
        ///                                                                   destinationMatrix: destination2
        ///                                                                 recurrentInputState: recurrent0
        ///                                                                recurrentOutputState: nil];
        /// ```
        ///
        /// # Safety
        ///
        /// - `source_offsets` must be a valid pointer or null.
        /// - `destination_offsets` must be a valid pointer or null.
        #[unsafe(method(encodeSequenceToCommandBuffer:sourceMatrices:sourceOffsets:destinationMatrices:destinationOffsets:recurrentInputState:recurrentOutputStates:))]
        #[unsafe(method_family = none)]
        pub unsafe fn encodeSequenceToCommandBuffer_sourceMatrices_sourceOffsets_destinationMatrices_destinationOffsets_recurrentInputState_recurrentOutputStates(
            &self,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            source_matrices: &NSArray<MPSMatrix>,
            source_offsets: *mut NSUInteger,
            destination_matrices: &NSArray<MPSMatrix>,
            destination_offsets: *mut NSUInteger,
            recurrent_input_state: Option<&MPSRNNRecurrentMatrixState>,
            recurrent_output_states: Option<&NSMutableArray<MPSRNNRecurrentMatrixState>>,
        );

        #[cfg(all(feature = "MPSMatrix", feature = "MPSState"))]
        #[unsafe(method(encodeSequenceToCommandBuffer:sourceMatrices:destinationMatrices:recurrentInputState:recurrentOutputStates:))]
        #[unsafe(method_family = none)]
        pub unsafe fn encodeSequenceToCommandBuffer_sourceMatrices_destinationMatrices_recurrentInputState_recurrentOutputStates(
            &self,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            source_matrices: &NSArray<MPSMatrix>,
            destination_matrices: &NSArray<MPSMatrix>,
            recurrent_input_state: Option<&MPSRNNRecurrentMatrixState>,
            recurrent_output_states: Option<&NSMutableArray<MPSRNNRecurrentMatrixState>>,
        );

        #[cfg(feature = "MPSMatrix")]
        /// Encode an MPSRNNMatrixInferenceLayer kernel stack for an input matrix sequences into a command buffer bidirectionally.
        /// The operation proceeds as follows: The first source matrix x0 is passed through all forward traversing layers in the stack,
        /// ie. those that were initialized with MPSRNNSequenceDirectionForward, recurrent input is assumed zero.
        /// This produces forward output yf0 and recurrent states hf00, hf01, hf02, ... hf0n, one for each forward layer in the stack.
        /// Then x1 is passed to forward layers together with recurrent state hf00, hf01, ..., hf0n, which produces yf1, and hf10,...
        /// This procedure is iterated until the last matrix in the input sequence x_(N-1), which produces forward output yf(N-1).
        /// The backwards layers iterate the same sequence backwards, starting from input x_(N-1) (recurrent state zero),
        /// that produces yb(N-1) and recurrent output hb(N-1)0, hf(N-1)1, ... hb(N-1)m, one for each backwards traversing layer.
        /// Then the backwards layers handle input x_(N-2) using recurrent state hb(N-1)0, ..., et cetera, until the
        /// first matrix of the sequence is computed, producing output yb0. The result of the operation is either pair of sequences
        /// ({yf0, yf1, ... , yf(N-1)},  {yb0, yb1, ... , yb(N-1)}) or a combined sequence, {(yf0 + yb0), ... , (yf(N-1) + yb(N-1)) },
        /// where '+' stands either for sum, or concatenation along feature channels, as specified by
        /// bidirectionalCombineMode.
        ///
        /// Parameter `commandBuffer`: A valid MTLCommandBuffer to receive the encoded filter
        ///
        /// Parameter `sourceSequence`: An array of valid MPSMatrix objects containing the source matrix sequence (x0, x1, ... x_n-1).
        ///
        /// Parameter `destinationForwardMatrices`: An array of valid MPSMatrices to be overwritten by result from forward input matrices. If bidirectionalCombineMode
        /// is either MPSRNNBidirectionalCombineModeAdd or MPSRNNBidirectionalCombineModeConcatenate, then will
        /// contain the combined results. destinationForwardMatrix may not alias with any of the source matrices.
        ///
        /// Parameter `destinationBackwardMatrices`: If bidirectionalCombineMode is MPSRNNBidirectionalCombineModeNone, then must be an array of valid MPSMatrices
        /// that will be overwritten by result from backward input matrices. Otherwise this parameter is ignored
        /// and can be nil. destinationBackwardMatrices may not alias to any of the source matrices.
        #[unsafe(method(encodeBidirectionalSequenceToCommandBuffer:sourceSequence:destinationForwardMatrices:destinationBackwardMatrices:))]
        #[unsafe(method_family = none)]
        pub unsafe fn encodeBidirectionalSequenceToCommandBuffer_sourceSequence_destinationForwardMatrices_destinationBackwardMatrices(
            &self,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            source_sequence: &NSArray<MPSMatrix>,
            destination_forward_matrices: &NSArray<MPSMatrix>,
            destination_backward_matrices: Option<&NSArray<MPSMatrix>>,
        );

        /// NSSecureCoding compatability
        ///
        /// See
        /// MPSKernel#initWithCoder.
        /// Parameter `aDecoder`: The NSCoder subclass with your serialized MPSRNNMatrixInferenceLayer
        ///
        /// Parameter `device`: The MTLDevice on which to make the MPSRNNMatrixInferenceLayer
        ///
        /// Returns: A new MPSRNNMatrixInferenceLayer object, or nil if failure.
        ///
        /// # Safety
        ///
        /// `a_decoder` possibly has further requirements.
        #[unsafe(method(initWithCoder:device:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithCoder_device(
            this: Allocated<Self>,
            a_decoder: &NSCoder,
            device: &ProtocolObject<dyn MTLDevice>,
        ) -> Option<Retained<Self>>;

        /// Make a copy of this kernel for a new device -
        ///
        /// See: MPSKernel
        ///
        /// Parameter `zone`: The NSZone in which to allocate the object
        ///
        /// Parameter `device`: The device for the new MPSKernel. If nil, then use
        /// self.device.
        ///
        /// Returns: a pointer to a copy of this MPSKernel. This will fail, returning
        /// nil if the device is not supported. Devices must be
        /// MTLFeatureSet_iOS_GPUFamily2_v1 or later.
        ///
        /// # Safety
        ///
        /// `zone` must be a valid pointer or null.
        #[unsafe(method(copyWithZone:device:))]
        #[unsafe(method_family = copy)]
        pub unsafe fn copyWithZone_device(
            &self,
            zone: *mut NSZone,
            device: Option<&ProtocolObject<dyn MTLDevice>>,
        ) -> Retained<Self>;
    );
}

/// Methods declared on superclass `MPSKernel`.
#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
impl MPSRNNMatrixInferenceLayer {
    extern_methods!(
        /// Called by NSCoder to decode MPSKernels
        ///
        /// This isn't the right interface to decode a MPSKernel, but
        /// it is the one that NSCoder uses. To enable your NSCoder
        /// (e.g. NSKeyedUnarchiver) to set which device to use
        /// extend the object to adopt the MPSDeviceProvider
        /// protocol. Otherwise, the Metal system default device
        /// will be used.
        ///
        /// # Safety
        ///
        /// `a_decoder` possibly has further requirements.
        #[unsafe(method(initWithCoder:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithCoder(
            this: Allocated<Self>,
            a_decoder: &NSCoder,
        ) -> Option<Retained<Self>>;
    );
}

/// Methods declared on superclass `NSObject`.
#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
impl MPSRNNMatrixInferenceLayer {
    extern_methods!(
        #[unsafe(method(init))]
        #[unsafe(method_family = init)]
        pub unsafe fn init(this: Allocated<Self>) -> Retained<Self>;

        #[unsafe(method(new))]
        #[unsafe(method_family = new)]
        pub unsafe fn new() -> Retained<Self>;
    );
}

extern_class!(
    /// Dependencies: This depends on Metal.framework
    ///
    /// This class holds the data that is passed from the forward pass needed in the backward pass.
    ///
    /// See also [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnnmatrixtrainingstate?language=objc)
    #[unsafe(super(MPSState, NSObject))]
    #[derive(Debug, PartialEq, Eq, Hash)]
    #[cfg(all(feature = "MPSCore", feature = "MPSState"))]
    pub struct MPSRNNMatrixTrainingState;
);

#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
extern_conformance!(
    unsafe impl NSObjectProtocol for MPSRNNMatrixTrainingState {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
impl MPSRNNMatrixTrainingState {
    extern_methods!();
}

/// Methods declared on superclass `MPSState`.
#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
impl MPSRNNMatrixTrainingState {
    extern_methods!(
        /// Create a MPSState holding a temporary MTLBuffer
        ///
        /// Parameter `cmdBuf`: The command buffer against which the temporary resource is allocated
        ///
        /// Parameter `bufferSize`: The size of the buffer in bytes
        #[unsafe(method(temporaryStateWithCommandBuffer:bufferSize:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer_bufferSize(
            cmd_buf: &ProtocolObject<dyn MTLCommandBuffer>,
            buffer_size: usize,
        ) -> Retained<Self>;

        /// Create a MPSState holding a temporary MTLTexture
        ///
        /// Parameter `cmdBuf`: The command buffer against which the temporary resource is allocated
        ///
        /// Parameter `descriptor`: A descriptor for the new temporary texture
        #[unsafe(method(temporaryStateWithCommandBuffer:textureDescriptor:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer_textureDescriptor(
            cmd_buf: &ProtocolObject<dyn MTLCommandBuffer>,
            descriptor: &MTLTextureDescriptor,
        ) -> Retained<Self>;

        /// Create a new autoreleased temporary state object without underlying resource
        ///
        /// Parameter `cmdBuf`: The command buffer with which the temporary resource is associated
        #[unsafe(method(temporaryStateWithCommandBuffer:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer(
            cmd_buf: &ProtocolObject<dyn MTLCommandBuffer>,
        ) -> Retained<Self>;

        #[unsafe(method(initWithDevice:bufferSize:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_bufferSize(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            buffer_size: usize,
        ) -> Retained<Self>;

        #[unsafe(method(initWithDevice:textureDescriptor:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_textureDescriptor(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            descriptor: &MTLTextureDescriptor,
        ) -> Retained<Self>;

        /// Create a MPSState with a non-temporary MTLResource
        ///
        /// Parameter `resource`: A MTLBuffer or MTLTexture. May be nil.
        ///
        /// # Safety
        ///
        /// - `resource` may need to be synchronized.
        /// - `resource` may be unretained, you must ensure it is kept alive while in use.
        #[unsafe(method(initWithResource:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithResource(
            this: Allocated<Self>,
            resource: Option<&ProtocolObject<dyn MTLResource>>,
        ) -> Retained<Self>;

        #[unsafe(method(init))]
        #[unsafe(method_family = init)]
        pub unsafe fn init(this: Allocated<Self>) -> Option<Retained<Self>>;

        /// Initialize a non-temporary state to hold a number of textures and buffers
        ///
        /// The allocation of each resource will be deferred  until it is needed.
        /// This occurs when -resource or -resourceAtIndex: is called.
        ///
        /// Parameter `resourceList`: The list of resources to create.
        #[unsafe(method(initWithDevice:resourceList:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_resourceList(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            resource_list: &MPSStateResourceList,
        ) -> Retained<Self>;

        /// Initialize a temporary state to hold a number of textures and buffers
        ///
        /// The textures occur first in sequence
        #[unsafe(method(temporaryStateWithCommandBuffer:resourceList:))]
        #[unsafe(method_family = none)]
        pub unsafe fn temporaryStateWithCommandBuffer_resourceList(
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            resource_list: &MPSStateResourceList,
        ) -> Retained<Self>;

        /// Create a state object with a list of MTLResources
        ///
        /// Because MPS prefers deferred allocation of resources
        /// your application should use -initWithTextures:bufferSizes:bufferCount:
        /// whenever possible. This method is useful for cases when the
        /// MTLResources must be initialized by the CPU.
        ///
        /// # Safety
        ///
        /// - `resources` generic may need to be synchronized.
        /// - `resources` generic may be unretained, you must ensure it is kept alive while in use.
        #[unsafe(method(initWithResources:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithResources(
            this: Allocated<Self>,
            resources: Option<&NSArray<ProtocolObject<dyn MTLResource>>>,
        ) -> Retained<Self>;
    );
}

/// Methods declared on superclass `NSObject`.
#[cfg(all(feature = "MPSCore", feature = "MPSState"))]
impl MPSRNNMatrixTrainingState {
    extern_methods!(
        #[unsafe(method(new))]
        #[unsafe(method_family = new)]
        pub unsafe fn new() -> Retained<Self>;
    );
}

/// [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnnmatrixid?language=objc)
// NS_ENUM
#[repr(transparent)]
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, PartialOrd, Ord)]
pub struct MPSRNNMatrixId(pub NSUInteger);
impl MPSRNNMatrixId {
    #[doc(alias = "MPSRNNMatrixIdSingleGateInputWeights")]
    pub const SingleGateInputWeights: Self = Self(0);
    #[doc(alias = "MPSRNNMatrixIdSingleGateRecurrentWeights")]
    pub const SingleGateRecurrentWeights: Self = Self(1);
    #[doc(alias = "MPSRNNMatrixIdSingleGateBiasTerms")]
    pub const SingleGateBiasTerms: Self = Self(2);
    #[doc(alias = "MPSRNNMatrixIdLSTMInputGateInputWeights")]
    pub const LSTMInputGateInputWeights: Self = Self(3);
    #[doc(alias = "MPSRNNMatrixIdLSTMInputGateRecurrentWeights")]
    pub const LSTMInputGateRecurrentWeights: Self = Self(4);
    #[doc(alias = "MPSRNNMatrixIdLSTMInputGateMemoryWeights")]
    pub const LSTMInputGateMemoryWeights: Self = Self(5);
    #[doc(alias = "MPSRNNMatrixIdLSTMInputGateBiasTerms")]
    pub const LSTMInputGateBiasTerms: Self = Self(6);
    #[doc(alias = "MPSRNNMatrixIdLSTMForgetGateInputWeights")]
    pub const LSTMForgetGateInputWeights: Self = Self(7);
    #[doc(alias = "MPSRNNMatrixIdLSTMForgetGateRecurrentWeights")]
    pub const LSTMForgetGateRecurrentWeights: Self = Self(8);
    #[doc(alias = "MPSRNNMatrixIdLSTMForgetGateMemoryWeights")]
    pub const LSTMForgetGateMemoryWeights: Self = Self(9);
    #[doc(alias = "MPSRNNMatrixIdLSTMForgetGateBiasTerms")]
    pub const LSTMForgetGateBiasTerms: Self = Self(10);
    #[doc(alias = "MPSRNNMatrixIdLSTMMemoryGateInputWeights")]
    pub const LSTMMemoryGateInputWeights: Self = Self(11);
    #[doc(alias = "MPSRNNMatrixIdLSTMMemoryGateRecurrentWeights")]
    pub const LSTMMemoryGateRecurrentWeights: Self = Self(12);
    #[doc(alias = "MPSRNNMatrixIdLSTMMemoryGateMemoryWeights")]
    pub const LSTMMemoryGateMemoryWeights: Self = Self(13);
    #[doc(alias = "MPSRNNMatrixIdLSTMMemoryGateBiasTerms")]
    pub const LSTMMemoryGateBiasTerms: Self = Self(14);
    #[doc(alias = "MPSRNNMatrixIdLSTMOutputGateInputWeights")]
    pub const LSTMOutputGateInputWeights: Self = Self(15);
    #[doc(alias = "MPSRNNMatrixIdLSTMOutputGateRecurrentWeights")]
    pub const LSTMOutputGateRecurrentWeights: Self = Self(16);
    #[doc(alias = "MPSRNNMatrixIdLSTMOutputGateMemoryWeights")]
    pub const LSTMOutputGateMemoryWeights: Self = Self(17);
    #[doc(alias = "MPSRNNMatrixIdLSTMOutputGateBiasTerms")]
    pub const LSTMOutputGateBiasTerms: Self = Self(18);
    #[doc(alias = "MPSRNNMatrixIdGRUInputGateInputWeights")]
    pub const GRUInputGateInputWeights: Self = Self(19);
    #[doc(alias = "MPSRNNMatrixIdGRUInputGateRecurrentWeights")]
    pub const GRUInputGateRecurrentWeights: Self = Self(20);
    #[doc(alias = "MPSRNNMatrixIdGRUInputGateBiasTerms")]
    pub const GRUInputGateBiasTerms: Self = Self(21);
    #[doc(alias = "MPSRNNMatrixIdGRURecurrentGateInputWeights")]
    pub const GRURecurrentGateInputWeights: Self = Self(22);
    #[doc(alias = "MPSRNNMatrixIdGRURecurrentGateRecurrentWeights")]
    pub const GRURecurrentGateRecurrentWeights: Self = Self(23);
    #[doc(alias = "MPSRNNMatrixIdGRURecurrentGateBiasTerms")]
    pub const GRURecurrentGateBiasTerms: Self = Self(24);
    #[doc(alias = "MPSRNNMatrixIdGRUOutputGateInputWeights")]
    pub const GRUOutputGateInputWeights: Self = Self(25);
    #[doc(alias = "MPSRNNMatrixIdGRUOutputGateRecurrentWeights")]
    pub const GRUOutputGateRecurrentWeights: Self = Self(26);
    #[doc(alias = "MPSRNNMatrixIdGRUOutputGateInputGateWeights")]
    pub const GRUOutputGateInputGateWeights: Self = Self(27);
    #[doc(alias = "MPSRNNMatrixIdGRUOutputGateBiasTerms")]
    pub const GRUOutputGateBiasTerms: Self = Self(28);
    #[doc(alias = "MPSRNNMatrixId_count")]
    pub const _count: Self = Self(29);
}

unsafe impl Encode for MPSRNNMatrixId {
    const ENCODING: Encoding = NSUInteger::ENCODING;
}

unsafe impl RefEncode for MPSRNNMatrixId {
    const ENCODING_REF: Encoding = Encoding::Pointer(&Self::ENCODING);
}

extern_class!(
    /// Dependencies: This depends on Metal.framework
    ///
    /// The MPSRNNMatrixTrainingLayer specifies a recurrent neural network layer for training on MPSMatrices.
    ///
    /// A MPSRNNMatrixTrainingLayer is initialized using a
    /// MPSRNNLayerDescriptor,which further specifies the
    /// recurrent network layer.
    /// The input and output vectors in encode calls are stored as rows of the input and output matrices and
    /// MPSRNNMatrixTrainingLayer supports matrices with decreasing number of rows: The row-indices identify the different
    /// sequences that may be of different lengths - for example if we have three sequences:
    /// ( x1, x2, x3 ), ( y1, y2, y3, y4 ) and ( z1, z2 )
    /// of vectors xi, yi and zi, then these can be inserted together as a batch to the sequence encoding kernel by
    /// using the matrices:
    ///
    /// ```text
    ///                            ( y1 )        ( y2 )        ( y3 )        ( y4 )
    ///                       m1 = ( x1 ),  m2 = ( x2 ),  m3 = ( x3 ),  m4 =
    ///                            ( z1 )        ( z2 )
    /// ```
    ///
    /// The gradient computation pass is then achieved by passing the corresponding gradient sequence from the
    /// previous layer ( dx1, dx2, dx3 ), ( dy1, dy2, dy3, dy4 ) and ( dz1, dz2 ) as matrices
    ///
    /// ```text
    ///                             ( dy1 )         ( dy2 )         ( dy3 )         ( dy4 )
    ///                       dm1 = ( dx1 ),  dm2 = ( dx2 ),  dm3 = ( dx3 ),  dm4 =
    ///                             ( dz1 )         ( dz2 )
    /// ```
    ///
    /// The mathematical operation described in the linear transformations of
    /// MPSRNNSingleGateDescriptorMPSLSTMDescriptorand
    /// MPSGRUDescriptorare y^T = W x^T
    /// <
    /// => y = x W^T, where x is the matrix containing
    /// the input vectors as rows, y is the matrix containing the output vectors as rows and W is the weight matrix.
    ///
    /// See also [Apple's documentation](https://developer.apple.com/documentation/metalperformanceshaders/mpsrnnmatrixtraininglayer?language=objc)
    #[unsafe(super(MPSKernel, NSObject))]
    #[derive(Debug, PartialEq, Eq, Hash)]
    #[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
    pub struct MPSRNNMatrixTrainingLayer;
);

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSCoding for MPSRNNMatrixTrainingLayer {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSCopying for MPSRNNMatrixTrainingLayer {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
unsafe impl CopyingHelper for MPSRNNMatrixTrainingLayer {
    type Result = Self;
}

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSObjectProtocol for MPSRNNMatrixTrainingLayer {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
extern_conformance!(
    unsafe impl NSSecureCoding for MPSRNNMatrixTrainingLayer {}
);

#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
impl MPSRNNMatrixTrainingLayer {
    extern_methods!(
        /// The number of feature channels input vector/matrix.
        #[unsafe(method(inputFeatureChannels))]
        #[unsafe(method_family = none)]
        pub unsafe fn inputFeatureChannels(&self) -> NSUInteger;

        /// The number of feature channels in the output vector/matrix.
        #[unsafe(method(outputFeatureChannels))]
        #[unsafe(method_family = none)]
        pub unsafe fn outputFeatureChannels(&self) -> NSUInteger;

        /// If YES then calls to functions
        /// encodeForwardSequenceToCommandBufferand
        /// encodeGradientSequenceToCommandBufferreturn every recurrent state
        /// in the array: recurrentOutputStates.
        /// Defaults to NO.
        #[unsafe(method(storeAllIntermediateStates))]
        #[unsafe(method_family = none)]
        pub unsafe fn storeAllIntermediateStates(&self) -> bool;

        /// Setter for [`storeAllIntermediateStates`][Self::storeAllIntermediateStates].
        #[unsafe(method(setStoreAllIntermediateStates:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setStoreAllIntermediateStates(&self, store_all_intermediate_states: bool);

        /// How recurrent output states from
        /// encodeForwardSequenceToCommandBufferand encodeGradientSequenceToCommandBuffer are constructed.
        /// Defaults to NO. For reference
        ///
        /// See: MPSState.
        #[unsafe(method(recurrentOutputIsTemporary))]
        #[unsafe(method_family = none)]
        pub unsafe fn recurrentOutputIsTemporary(&self) -> bool;

        /// Setter for [`recurrentOutputIsTemporary`][Self::recurrentOutputIsTemporary].
        #[unsafe(method(setRecurrentOutputIsTemporary:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setRecurrentOutputIsTemporary(&self, recurrent_output_is_temporary: bool);

        /// How training output states from
        /// encodeForwardSequenceToCommandBufferare constructed.
        /// Defaults to NO. For reference
        ///
        /// See: MPSState.
        #[unsafe(method(trainingStateIsTemporary))]
        #[unsafe(method_family = none)]
        pub unsafe fn trainingStateIsTemporary(&self) -> bool;

        /// Setter for [`trainingStateIsTemporary`][Self::trainingStateIsTemporary].
        #[unsafe(method(setTrainingStateIsTemporary:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setTrainingStateIsTemporary(&self, training_state_is_temporary: bool);

        /// If yes then the computed weight gradients are accumulated on top of existing values in
        /// calls to the gradient computation functions: encodeGradientSequenceToCommandBuffer.
        /// Defaults to NO.
        #[unsafe(method(accumulateWeightGradients))]
        #[unsafe(method_family = none)]
        pub unsafe fn accumulateWeightGradients(&self) -> bool;

        /// Setter for [`accumulateWeightGradients`][Self::accumulateWeightGradients].
        #[unsafe(method(setAccumulateWeightGradients:))]
        #[unsafe(method_family = none)]
        pub unsafe fn setAccumulateWeightGradients(&self, accumulate_weight_gradients: bool);

        #[cfg(feature = "MPSMatrix")]
        /// Initializes a linear (fully connected) RNN kernel for training
        ///
        /// Parameter `device`: The MTLDevice on which this MPSRNNMatrixLayer filter will be used
        ///
        /// Parameter `rnnDescriptor`: The descriptor that defines the RNN layer
        ///
        /// Parameter `trainableWeights`: An array where to store the weights of the layer as MPSMatrices.
        /// NOTE: The exact layout and number of matrices may vary between
        /// platforms and therefore you should not save out these weights directly,
        /// but instead use the function encodeCopyWeightsToCommandBuffer to identify
        /// the weights and biases for serialization.
        /// Typically you should pass here an initialized but empty NSMutableArray and
        /// when this function returns the array will have been populated with the
        /// weight matrices needed in the encode-calls, by using initial values from
        /// the datasources in rnnDescriptor.
        ///
        /// Returns: A valid MPSRNNMatrixTrainingLayer object or nil, if failure.
        #[unsafe(method(initWithDevice:rnnDescriptor:trainableWeights:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice_rnnDescriptor_trainableWeights(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
            rnn_descriptor: &MPSRNNDescriptor,
            trainable_weights: &NSMutableArray<MPSMatrix>,
        ) -> Retained<Self>;

        #[cfg(all(feature = "MPSCoreTypes", feature = "MPSMatrix"))]
        /// Initializes a set of matrices that can be used in training for weight and bias gradient outputs in
        ///
        /// See: encodeBackwardSequenceToCommandBuffer. Can be also used to easily create auxiliary matrices for example
        /// for ADAM and other advanced optimization schemes. The layout and number of matrices is the same as for the outputs of
        ///
        /// See: initWithDevice, but the data type may differ. NOTE: These matrices cannot be used as weight matrices in the
        /// forward and backward encode calls, but matrices from initWithDevice() or createWeightMatrices() should be used instead.
        ///
        /// Parameter `matricesOut`: An array where the newly created matrices will be stored, will be initialized to zero.
        ///
        /// Parameter `dataType`: Datatype for the entries - currently MPSDataTypeFloat32 and MPSDataTypeFloat16 are supported.
        #[unsafe(method(createWeightGradientMatrices:dataType:))]
        #[unsafe(method_family = none)]
        pub unsafe fn createWeightGradientMatrices_dataType(
            &self,
            matrices_out: &NSMutableArray<MPSMatrix>,
            data_type: MPSDataType,
        );

        #[cfg(all(feature = "MPSCoreTypes", feature = "MPSMatrix"))]
        /// As
        /// createWeightGradientMatrices,but the matrices will be temporary with readCount = 1, which means that they
        /// become invalid after the first encode call that reads them. Note also that as the matrices are temporary, their
        /// storage mode will be private which means that you can only access the data using a kernel on the GPU.
        ///
        /// Parameter `matricesOut`: An array where the newly created matrices will be stored, will be initialized to zero.
        ///
        /// Parameter `dataType`: Datatype for the entries - currently MPSDataTypeFloat32 and MPSDataTypeFloat16 are supported.
        ///
        /// Parameter `commandBuffer`: The command buffer that the temporary matrices will live on.
        #[unsafe(method(createTemporaryWeightGradientMatrices:dataType:commandBuffer:))]
        #[unsafe(method_family = none)]
        pub unsafe fn createTemporaryWeightGradientMatrices_dataType_commandBuffer(
            &self,
            matrices_out: &NSMutableArray<MPSMatrix>,
            data_type: MPSDataType,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
        );

        #[cfg(feature = "MPSMatrix")]
        /// Initializes a set of matrices that can be used in training for weight and bias matrices in
        /// the forward and backward passes. The layout, datatype and number of matrices is the same as for the outputs of
        ///
        /// See: initWithDevice.
        ///
        /// Parameter `matricesOut`: An array where the newly created matrices will be stored, will be initialized to zero.
        #[unsafe(method(createWeightMatrices:))]
        #[unsafe(method_family = none)]
        pub unsafe fn createWeightMatrices(&self, matrices_out: &NSMutableArray<MPSMatrix>);

        #[unsafe(method(initWithDevice:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithDevice(
            this: Allocated<Self>,
            device: &ProtocolObject<dyn MTLDevice>,
        ) -> Retained<Self>;

        #[cfg(feature = "MPSMatrix")]
        /// Encode a copy kernel that copies one matrix from the trainable weight set to a matrix with standard layout,
        /// where the column index is the input feature channel index (in forward direction) and row index is the output
        /// feature channel index.
        ///
        /// Parameter `commandBuffer`: A valid MTLCommandBuffer to receive the encoded filter
        ///
        /// Parameter `weights`: An array weights from
        ///
        /// See: initWithDevice or
        ///
        /// See: createWeightMatrices.
        ///
        /// Parameter `matrixId`: Which matrix to copy - has to be a valid Id based on inputs defined in
        /// the rnnDescriptor of
        ///
        /// See: initWithDevice.
        ///
        /// Parameter `matrix`: The destination or source matrix that is used in the copy.
        ///
        /// Parameter `copyFromWeightsToMatrix`: If YES then the copy direction is from the set of trainable 'weights' to 'matrix',
        /// otherwise the copy is done from 'matrix' to 'weights'.
        ///
        /// Parameter `matrixOffset`: A (valid) offset into matrix to be applied to the copy operation.
        #[unsafe(method(encodeCopyWeightsToCommandBuffer:weights:matrixId:matrix:copyFromWeightsToMatrix:matrixOffset:))]
        #[unsafe(method_family = none)]
        pub unsafe fn encodeCopyWeightsToCommandBuffer_weights_matrixId_matrix_copyFromWeightsToMatrix_matrixOffset(
            &self,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            weights: &NSArray<MPSMatrix>,
            matrix_id: MPSRNNMatrixId,
            matrix: &MPSMatrix,
            copy_from_weights_to_matrix: bool,
            matrix_offset: MTLOrigin,
        );

        #[cfg(all(feature = "MPSMatrix", feature = "MPSState"))]
        /// Encode an MPSRNNMatrixTrainingLayer forward pass kernel for a sequence of inputs into a command buffer.
        ///
        /// Parameter `commandBuffer`: A valid MTLCommandBuffer to receive the encoded filter
        ///
        /// Parameter `sourceMatrices`: An array of valid MPSMatrix objects containing the sequence of source matrices.
        ///
        /// Parameter `sourceOffsets`: An array of byte-offsets into the sourceMatrices, if nil zeros are assumed and
        /// if not nil must contain offset for every matrix in sourceMatrices.
        ///
        /// Parameter `destinationMatrices`: An array valid MPSMatrices to be overwritten by result matrix sequence.
        /// destinationMatrices may not alias sourceMatrices.
        ///
        /// Parameter `destinationOffsets`: An array of byte-offsets into the destinationMatrices, if nil zeros are assumed and
        /// if not nil must contain offset for every matrix in destinationMatrices.
        ///
        /// Parameter `trainingStates`: An array containing the training states to be passed to the gradient computation
        /// encode function.
        ///
        /// Parameter `recurrentInputState`: An optional state containing the output matrices and memory cells (for LSTMs)
        /// of the layer obtained from the previous input matrices in a sequence of inputs.
        /// Has to be the output of a previous call to this function or nil (assumed zero).
        ///
        /// Parameter `recurrentOutputStates`: An array that will be appended with the recurrent output states. May not be nil.
        /// If recurrentOutputIsTemporary is YES and then all returned recurrent states
        /// will be temporary.
        ///
        /// See: MPSState:isTemporary.
        ///
        /// Parameter `weights`: An array of valid MPSMatrix objects containing the weights, should be the array
        /// that was produced either by
        ///
        /// See: initWithDevice or
        ///
        /// See: createWeightMatrices.
        ///
        /// # Safety
        ///
        /// - `source_offsets` must be a valid pointer or null.
        /// - `destination_offsets` must be a valid pointer or null.
        #[unsafe(method(encodeForwardSequenceToCommandBuffer:sourceMatrices:sourceOffsets:destinationMatrices:destinationOffsets:trainingStates:recurrentInputState:recurrentOutputStates:weights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn encodeForwardSequenceToCommandBuffer_sourceMatrices_sourceOffsets_destinationMatrices_destinationOffsets_trainingStates_recurrentInputState_recurrentOutputStates_weights(
            &self,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            source_matrices: &NSArray<MPSMatrix>,
            source_offsets: *mut NSUInteger,
            destination_matrices: &NSArray<MPSMatrix>,
            destination_offsets: *mut NSUInteger,
            training_states: &NSMutableArray<MPSRNNMatrixTrainingState>,
            recurrent_input_state: Option<&MPSRNNRecurrentMatrixState>,
            recurrent_output_states: Option<&NSMutableArray<MPSRNNRecurrentMatrixState>>,
            weights: &NSArray<MPSMatrix>,
        );

        #[cfg(all(feature = "MPSMatrix", feature = "MPSState"))]
        /// Encode an MPSRNNMatrixTrainingLayer forward pass kernel for a sequence of inputs into a command buffer.
        ///
        /// Parameter `commandBuffer`: A valid MTLCommandBuffer to receive the encoded filter
        ///
        /// Parameter `sourceMatrices`: An array of valid MPSMatrix objects containing the sequence of source matrices.
        ///
        /// Parameter `destinationMatrices`: An array valid MPSMatrices to be overwritten by result matrix sequence.
        /// destinationMatrices may not alias sourceMatrices.
        ///
        /// Parameter `trainingStates`: An array containing the training states to be passed to the gradient computation
        /// encode function.
        ///
        /// Parameter `weights`: An array of valid MPSMatrix objects containing the weights, should be the array
        /// that was produced either by
        ///
        /// See: initWithDevice or
        ///
        /// See: createWeightMatrices.
        #[unsafe(method(encodeForwardSequenceToCommandBuffer:sourceMatrices:destinationMatrices:trainingStates:weights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn encodeForwardSequenceToCommandBuffer_sourceMatrices_destinationMatrices_trainingStates_weights(
            &self,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            source_matrices: &NSArray<MPSMatrix>,
            destination_matrices: &NSArray<MPSMatrix>,
            training_states: &NSMutableArray<MPSRNNMatrixTrainingState>,
            weights: &NSArray<MPSMatrix>,
        );

        #[cfg(all(feature = "MPSMatrix", feature = "MPSState"))]
        /// Encode an MPSRNNMatrixTrainingLayer gradient pass kernel for a sequence of input gradients into a command buffer.
        /// NOTE: The time sequence indexing follows the array indexing in the inputs: sourceGradients[0] has to contain the
        /// gradients corresponding to the first matrix in the forward pass corresponding to the current subsequence, which is
        /// typically sourceMatrices[0].
        ///
        /// Parameter `commandBuffer`: A valid MTLCommandBuffer to receive the encoded filter
        ///
        /// Parameter `forwardSources`: An array of MPSMatrix objects containing the sequence of source matrices of the forward pass.
        ///
        /// Parameter `forwardSourceOffsets`: An array of byte-offsets into the forwardSources, if nil zeros are assumed and
        /// if not nil must contain offset for every matrix in forwardSources.
        ///
        /// Parameter `sourceGradients`: An array of valid MPSMatrix objects containing the sequence of source gradient matrices.
        ///
        /// Parameter `sourceGradientOffsets`: An array of byte-offsets into the sourceGradients, if nil zeros are assumed and
        /// if not nil must contain offset for every matrix in sourceGradients.
        ///
        /// Parameter `destinationGradients`: An array valid MPSMatrix objects that will receive the backpropagated gradients, may be
        /// nil if not needed (for example first layer in graph).
        ///
        /// Parameter `destinationOffsets`: An array of byte-offsets into the destinationGradients, if nil zeros are assumed and
        /// if not nil must contain offset for every matrix in destinationGradients.
        ///
        /// Parameter `weightGradients`: An array of valid MPSMatrix objects that will receive the gradient wrt. weights and
        /// biases of the layer - should be the array that was produced either
        /// by
        ///
        /// See: initWithDevice or
        ///
        /// See: createWeightMatrices. May be nil in which case
        /// the gradients for the weights are not computed.
        ///
        /// Parameter `trainingStates`: An array containing the training states from the forward pass - the array must contain
        /// the states corresponding to the input gradients is sourceGradients.
        ///
        /// Parameter `recurrentInputState`: An optional state containing the output matrices and memory cells (for LSTMs)
        /// of the layer obtained from the previous input gradients in a sequence of inputs.
        /// Has to be the output of a previous call to this function or nil (assumed zero).
        ///
        /// Parameter `recurrentOutputStates`: An array that will be appended with the recurrent output states. Can be nil.
        /// If recurrentOutputIsTemporary is YES and then all returned recurrent states
        /// will be temporary.
        ///
        /// See: MPSState:isTemporary.
        ///
        /// Parameter `weights`: An array of valid MPSMatrix objects containing the weights, should be the array
        /// that was produced either by
        ///
        /// See: initWithDevice or
        ///
        /// See: createWeightMatrices.
        ///
        /// # Safety
        ///
        /// - `forward_source_offsets` must be a valid pointer or null.
        /// - `source_gradient_offsets` must be a valid pointer or null.
        /// - `destination_offsets` must be a valid pointer or null.
        #[unsafe(method(encodeGradientSequenceToCommandBuffer:forwardSources:forwardSourceOffsets:sourceGradients:sourceGradientOffsets:destinationGradients:destinationOffsets:weightGradients:trainingStates:recurrentInputState:recurrentOutputStates:weights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn encodeGradientSequenceToCommandBuffer_forwardSources_forwardSourceOffsets_sourceGradients_sourceGradientOffsets_destinationGradients_destinationOffsets_weightGradients_trainingStates_recurrentInputState_recurrentOutputStates_weights(
            &self,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            forward_sources: &NSArray<MPSMatrix>,
            forward_source_offsets: *mut NSUInteger,
            source_gradients: &NSArray<MPSMatrix>,
            source_gradient_offsets: *mut NSUInteger,
            destination_gradients: Option<&NSArray<MPSMatrix>>,
            destination_offsets: *mut NSUInteger,
            weight_gradients: Option<&NSArray<MPSMatrix>>,
            training_states: &NSArray<MPSRNNMatrixTrainingState>,
            recurrent_input_state: Option<&MPSRNNRecurrentMatrixState>,
            recurrent_output_states: Option<&NSMutableArray<MPSRNNRecurrentMatrixState>>,
            weights: &NSArray<MPSMatrix>,
        );

        #[cfg(all(feature = "MPSMatrix", feature = "MPSState"))]
        /// Encode an MPSRNNMatrixTrainingLayer gradient pass kernel for a sequence of input gradients into a command buffer.
        /// NOTE: The time sequence indexing follows the array indexing in the inputs: sourceGradients[0] has to contain the
        /// gradients corresponding to the first matrix in the forward pass corresponding to the current subsequence, which is
        /// typically sourceMatrices[0].
        ///
        /// Parameter `commandBuffer`: A valid MTLCommandBuffer to receive the encoded filter
        ///
        /// Parameter `forwardSources`: An array of MPSMatrix objects containing the sequence of source matrices of the forward pass.
        ///
        /// Parameter `sourceGradients`: An array of MPSMatrix objects containing the sequence of source gradient matrices.
        ///
        /// Parameter `destinationGradients`: An array valid MPSMatrix objects that will receive the backpropagated gradients, may be
        /// nil if not needed (for example first layer in graph).
        ///
        /// Parameter `weightGradients`: An array valid MPSMatrix objects that will receive the gradient wrt. weights and
        /// biases of the layer - should be the array that was produced either
        /// by
        ///
        /// See: initWithDevice or
        ///
        /// See: createWeightMatrices. May be nil in which case
        /// the gradients for the weights are not computed.
        /// NOTE: The weight gradients are accumulated on top of existing values so
        ///
        ///
        /// Parameter `trainingStates`: An array containing the training states from the forward pass - the array must contain
        /// the states corresponding to the input gradients is sourceGradients.
        ///
        /// Parameter `weights`: An array of valid MPSMatrix objects containing the weights, should be the array
        /// that was produced either by
        ///
        /// See: initWithDevice or
        ///
        /// See: createWeightMatrices.
        #[unsafe(method(encodeGradientSequenceToCommandBuffer:forwardSources:sourceGradients:destinationGradients:weightGradients:trainingStates:weights:))]
        #[unsafe(method_family = none)]
        pub unsafe fn encodeGradientSequenceToCommandBuffer_forwardSources_sourceGradients_destinationGradients_weightGradients_trainingStates_weights(
            &self,
            command_buffer: &ProtocolObject<dyn MTLCommandBuffer>,
            forward_sources: &NSArray<MPSMatrix>,
            source_gradients: &NSArray<MPSMatrix>,
            destination_gradients: Option<&NSArray<MPSMatrix>>,
            weight_gradients: Option<&NSArray<MPSMatrix>>,
            training_states: &NSArray<MPSRNNMatrixTrainingState>,
            weights: &NSArray<MPSMatrix>,
        );

        /// NSSecureCoding compatability
        ///
        /// See
        /// MPSKernel#initWithCoder.
        /// Parameter `aDecoder`: The NSCoder subclass with your serialized MPSRNNMatrixTrainingLayer
        ///
        /// Parameter `device`: The MTLDevice on which to make the MPSRNNMatrixTrainingLayer
        ///
        /// Returns: A new MPSRNNMatrixTrainingLayer object, or nil if failure.
        ///
        /// # Safety
        ///
        /// `a_decoder` possibly has further requirements.
        #[unsafe(method(initWithCoder:device:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithCoder_device(
            this: Allocated<Self>,
            a_decoder: &NSCoder,
            device: &ProtocolObject<dyn MTLDevice>,
        ) -> Option<Retained<Self>>;

        /// Make a copy of this kernel for a new device -
        ///
        /// See: MPSKernel
        ///
        /// Parameter `zone`: The NSZone in which to allocate the object
        ///
        /// Parameter `device`: The device for the new MPSKernel. If nil, then use
        /// self.device.
        ///
        /// Returns: a pointer to a copy of this MPSKernel. This will fail, returning
        /// nil if the device is not supported. Devices must be
        /// MTLFeatureSet_iOS_GPUFamily2_v1 or later.
        ///
        /// # Safety
        ///
        /// `zone` must be a valid pointer or null.
        #[unsafe(method(copyWithZone:device:))]
        #[unsafe(method_family = copy)]
        pub unsafe fn copyWithZone_device(
            &self,
            zone: *mut NSZone,
            device: Option<&ProtocolObject<dyn MTLDevice>>,
        ) -> Retained<Self>;
    );
}

/// Methods declared on superclass `MPSKernel`.
#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
impl MPSRNNMatrixTrainingLayer {
    extern_methods!(
        /// Called by NSCoder to decode MPSKernels
        ///
        /// This isn't the right interface to decode a MPSKernel, but
        /// it is the one that NSCoder uses. To enable your NSCoder
        /// (e.g. NSKeyedUnarchiver) to set which device to use
        /// extend the object to adopt the MPSDeviceProvider
        /// protocol. Otherwise, the Metal system default device
        /// will be used.
        ///
        /// # Safety
        ///
        /// `a_decoder` possibly has further requirements.
        #[unsafe(method(initWithCoder:))]
        #[unsafe(method_family = init)]
        pub unsafe fn initWithCoder(
            this: Allocated<Self>,
            a_decoder: &NSCoder,
        ) -> Option<Retained<Self>>;
    );
}

/// Methods declared on superclass `NSObject`.
#[cfg(all(feature = "MPSCore", feature = "MPSKernel"))]
impl MPSRNNMatrixTrainingLayer {
    extern_methods!(
        #[unsafe(method(init))]
        #[unsafe(method_family = init)]
        pub unsafe fn init(this: Allocated<Self>) -> Retained<Self>;

        #[unsafe(method(new))]
        #[unsafe(method_family = new)]
        pub unsafe fn new() -> Retained<Self>;
    );
}