DirectGpuTensorEngine

Models & Types · AiDotNet.Tensors.Engines

No summary documentation available yet.

Inheritance: CpuEngine → DirectGpuTensorEngine

Implements: IDisposable, IInferenceGraphCaptureEngine

Known derived types: GpuEngine

Constructors

Constructor Summary
DirectGpuTensorEngine()
DirectGpuTensorEngine(DirectGpuEngine directGpu)

Properties

Property Summary
GpuExecutionContext CurrentContext { get; }
bool IsGpuAvailable { get; }
bool IsGpuContextActive { get; }
bool RecordsFusedLstmSequenceTraining { get; }
bool SupportsDeferredExecution { get; }
bool SupportsGpu { get; }
bool SupportsInferenceGraphCapture { get; }
int CachedTensorCount { get; }
string Name { get; }

Methods

Method Summary
GpuExecutionContext BeginGpuContext(GpuExecutionOptions options)
GpuScope BeginGpuScope()
GpuStreamPool CreateStreamPool(GpuExecutionOptions options)
GpuStreamScheduler GetStreamScheduler(GpuStreamPool streamPool, GpuStreamType streamType)
IAsyncGpuBackend GetAsyncBackend()
IDeferredScope BeginDeferredScope(GpuExecutionOptions options)
IDirectGpuBackend GetBackend()
IDisposable EnterResidentCaptureScope()
IntPtr CaptureGpuGraph(Action forward)
SparseTensor<T> CreateCsrFromDense(Tensor<T> denseTensor, float threshold)
SparseTensor<T> CreateCsrFromEdges(int[] sourceIndices, int[] targetIndices, float[] values, int numNodes)
T TensorMedian(Tensor<T> input)
T TensorNanMedian(Tensor<T> input)
T TensorTrace(Tensor<T> tensor)
T TensorVecDot(Tensor<T> a, Tensor<T> b)
TResult WithGpuContext(Func<GpuExecutionContext, TResult> func, Func<TResult> fallback, GpuExecutionOptions options)
Tensor<Bit> TensorEq(Tensor<T> a, Tensor<T> b)
Tensor<Bit> TensorEqScalar(Tensor<T> a, T scalar)
Tensor<Bit> TensorIsClose(Tensor<T> a, Tensor<T> b, T rtol, T atol, bool equalNan)
Tensor<Bit> TensorIsFinite(Tensor<T> tensor)
Tensor<Bit> TensorIsIn(Tensor<T> elements, Tensor<T> testElements, bool invert)
Tensor<Bit> TensorIsInf(Tensor<T> tensor)
Tensor<Bit> TensorIsNan(Tensor<T> tensor)
Tensor<Bit> TensorLogicalAnd(Tensor<Bit> a, Tensor<Bit> b)
Tensor<Bit> TensorLogicalNot(Tensor<Bit> a)
Tensor<Bit> TensorLogicalOr(Tensor<Bit> a, Tensor<Bit> b)
Tensor<Bit> TensorLogicalXor(Tensor<Bit> a, Tensor<Bit> b)
Tensor<Complex<T>> NativeAnalyticSignal(Tensor<T> input, double freqLow, double freqHigh, double sampleRate)
Tensor<Complex<T>> NativeBispectrum(Tensor<Complex<T>> spectrum, int maxF1, int maxF2)
Tensor<Complex<T>> NativeComplexAdd(Tensor<Complex<T>> a, Tensor<Complex<T>> b)
Tensor<Complex<T>> NativeComplexConjugate(Tensor<Complex<T>> a)
Tensor<Complex<T>> NativeComplexCrossSpectral(Tensor<Complex<T>> x, Tensor<Complex<T>> y)
Tensor<Complex<T>> NativeComplexFFT(Tensor<T> input)
Tensor<Complex<T>> NativeComplexFFT2D(Tensor<T> input)
Tensor<Complex<T>> NativeComplexFFTComplex(Tensor<Complex<T>> input)
Tensor<Complex<T>> NativeComplexFFTND(Tensor<T> input, int[] axes)
Tensor<Complex<T>> NativeComplexFromPolar(Tensor<T> magnitudes, Tensor<T> phases)
Tensor<Complex<T>> NativeComplexIFFT(Tensor<Complex<T>> input)
Tensor<Complex<T>> NativeComplexMultiply(Tensor<Complex<T>> a, Tensor<Complex<T>> b)
Tensor<Complex<T>> NativeComplexScale(Tensor<Complex<T>> a, T scalar)
Tensor<Complex<T>> NativeComplexTopK(Tensor<Complex<T>> input, int k)
Tensor<Complex<T>> NativeTrispectrum(Tensor<Complex<T>> spectrum, int maxF1, int maxF2, int maxF3)
Tensor<Complex<T>> TensorComplex(Tensor<T> real, Tensor<T> imag)
Tensor<Complex<T>> TensorViewAsComplex(Tensor<T> input)
Tensor<T> AbcScanForward(Tensor<T> qProj, Tensor<T> kProj, Tensor<T> vProj, Tensor<T> forgetGate, Tensor<T> slotKeys, int numHeads, double slotInitScale)
Tensor<T> ActivationGpu(Tensor<T> input, FusedActivationType activation)
Tensor<T> AdaptiveAvgPool2D(Tensor<T> input, int outputHeight, int outputWidth)
Tensor<T> AddBiasGpu(Tensor<T> input, Tensor<T> bias)
Tensor<T> AddGpu(Tensor<T> a, Tensor<T> b)
Tensor<T> AffineGrid3D(Tensor<T> theta, int outputDepth, int outputHeight, int outputWidth, bool alignCorners)
Tensor<T> AffineGridGpu(Tensor<T> theta, int batch, int outputHeight, int outputWidth)
Tensor<T> AmplitudeToDB(Tensor<T> input, float minAmplitude, float? topDb)
Tensor<T> ApplyRoPEInterleaved(Tensor<T> input, Tensor<T> cos, Tensor<T> sin, int startPosition)
Tensor<T> ApplyRoPEInterleavedGpu(Tensor<T> input, Tensor<T> cos, Tensor<T> sin, int startPosition)
Tensor<T> ArgMaxAxisGpu(Tensor<T> input, int axis)
Tensor<T> AvgPool2D(Tensor<T> input, int poolSize, int stride, int padding, bool countIncludePad)
Tensor<T> AvgPool2D(Tensor<T> input, int[] poolSize, int[] stride)
Tensor<T> AvgPool2DBackward(Tensor<T> gradOutput, int[] inputShape, int[] poolSize, int[] stride, int[] padding, bool countIncludePad)
Tensor<T> AvgPool2DBackwardGpu(Tensor<T> gradOutput, int[] inputShape, int[] poolSize, int[] stride)
Tensor<T> AvgPool2DGpu(Tensor<T> input, int[] poolSize, int[] stride)
Tensor<T> BatchMatMul(Tensor<T> a, Tensor<T> b)
Tensor<T> BatchNorm(Tensor<T> input, Tensor<T> gamma, Tensor<T> beta, double epsilon, Tensor<T> mean, Tensor<T> variance)
Tensor<T> BatchNormActivationConv2D(Tensor<T> input, Tensor<T> preGamma, Tensor<T> preBeta, Tensor<T> preMean, Tensor<T> preVariance, double preEpsilon, Tensor<T> kernel, Tensor<T> bias, int strideH, int strideW, int padH, int padW, int dilationH, int dilationW, FusedActivationType activation)
Tensor<T> BatchNormActivationConv2DBatchNorm(Tensor<T> input, Tensor<T> preGamma, Tensor<T> preBeta, Tensor<T> preMean, Tensor<T> preVariance, double preEpsilon, Tensor<T> kernel, Tensor<T> bias, Tensor<T> gamma, Tensor<T> beta, Tensor<T> runningMean, Tensor<T> runningVariance, double epsilon, int strideH, int strideW, int padH, int padW, int dilationH, int dilationW, FusedActivationType activation)
Tensor<T> BatchNormAffine(Tensor<T> x, Tensor<T> gamma, Tensor<T> beta, Tensor<T> mean, Tensor<T> variance, double epsilon)
Tensor<T> BatchNormInference(Tensor<T> x, Tensor<T> gamma, Tensor<T> beta, Tensor<T> mean, Tensor<T> variance, double epsilon)
Tensor<T> BatchNormInference(Tensor<T> x, Tensor<T> gamma, Tensor<T> beta, Tensor<T> mean, Tensor<T> variance, double epsilon, FusedActivationType activation)
Tensor<T> BatchedMatMulGpu(Tensor<T> input, Tensor<T> weights)
Tensor<T> BoxArea(Tensor<T> boxes)
Tensor<T> BoxConvert(Tensor<T> boxes, BoxFormat from, BoxFormat to)
Tensor<T> BoxIou(Tensor<T> boxesA, Tensor<T> boxesB)
Tensor<T> BroadcastAddGpu(Tensor<T> a, Tensor<T> b)
Tensor<T> BroadcastMultiplyColumnGpu(Tensor<T> input, Tensor<T> weights)
Tensor<T> BroadcastMultiplyRowGpu(Tensor<T> input, Tensor<T> weights)
Tensor<T> CapsulePredictionsGpu(Tensor<T> input, IGpuBuffer weightsBuffer, int outputCapsules, int outputDim)
Tensor<T> CapsulePredictionsGpu(Tensor<T> input, Tensor<T> weights)
Tensor<T> CapsuleRoutingGpu(Tensor<T> transformedInput, Tensor<T> bias, int numIterations, float epsilon)
Tensor<T> CapsuleTransformGpu(Tensor<T> input, IGpuBuffer weightsBuffer, int numCapsules, int capsuleDim)
Tensor<T> CapsuleTransformGpu(Tensor<T> input, Tensor<T> weights)
Tensor<T> CausalDepthwiseConv1D(Tensor<T> input, Tensor<T> kernel)
Tensor<T> CausalDepthwiseConv1DBackwardInput(Tensor<T> gradOutput, Tensor<T> kernel)
Tensor<T> CausalDepthwiseConv1DBackwardKernel(Tensor<T> gradOutput, Tensor<T> input, int taps)
Tensor<T> CompleteBoxIou(Tensor<T> boxesA, Tensor<T> boxesB)
Tensor<T> ComplexDiagonalSsmScanForward(Tensor<T> input, Tensor<T> transitionReal, Tensor<T> transitionImag, Tensor<T> inputMapReal, Tensor<T> inputMapImag, Tensor<T> outputMapReal, Tensor<T> outputMapImag, Tensor<T> skip)
Tensor<T> ComputeDeltas(Tensor<T> input, int winLength)
Tensor<T> ConcatGpu(Tensor<T>[] inputs, int axis)
Tensor<T> Conv1D(Tensor<T> input, Tensor<T> kernel, int stride, int padding, int dilation)
Tensor<T> Conv2D(Tensor<T> input, Tensor<T> kernel, int stride, int padding, int dilation)
Tensor<T> Conv2D(Tensor<T> input, Tensor<T> kernel, int[] stride, int[] padding, int[] dilation)
Tensor<T> Conv2DBackwardBiasGpu(Tensor<T> gradOutput)
Tensor<T> Conv2DBackwardInputGpu(Tensor<T> gradOutput, Tensor<T> kernel, int[] inputShape, int[] stride, int[] padding, int[] dilation)
Tensor<T> Conv2DBackwardKernelGpu(Tensor<T> gradOutput, Tensor<T> input, int[] kernelShape, int[] stride, int[] padding, int[] dilation)
Tensor<T> Conv3D(Tensor<T> input, Tensor<T> kernel, int stride, int padding, int dilation)
Tensor<T> ConvTranspose2D(Tensor<T> input, Tensor<T> kernel, int[] stride, int[] padding, int[] outputPadding)
Tensor<T> ConvTranspose2DBackwardInputGpu(Tensor<T> gradOutput, Tensor<T> kernel, int[] inputShape, int[] stride, int[] padding, int[] outputPadding)
Tensor<T> ConvTranspose2DBackwardKernelGpu(Tensor<T> gradOutput, Tensor<T> input, int[] kernelShape, int[] stride, int[] padding, int[] outputPadding)
Tensor<T> CosGpu(Tensor<T> input)
Tensor<T> DeformableConv2D(Tensor<T> input, Tensor<T> kernel, Tensor<T> offsets, Tensor<T> mask, int[] stride, int[] padding, int[] dilation)
Tensor<T> DeformableConv2DBackwardInput(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> kernel, Tensor<T> offsets, Tensor<T> mask, int[] inputShape, int[] stride, int[] padding, int[] dilation)
Tensor<T> DeformableConv2DBackwardKernel(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> offsets, Tensor<T> mask, int[] kernelShape, int[] stride, int[] padding, int[] dilation)
Tensor<T> DeformableConv2DBackwardMask(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> kernel, Tensor<T> offsets, Tensor<T> mask, int[] stride, int[] padding, int[] dilation)
Tensor<T> DeformableConv2DBackwardOffset(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> kernel, Tensor<T> offsets, Tensor<T> mask, int[] stride, int[] padding, int[] dilation)
Tensor<T> DeformableConv2DGpu(Tensor<T> input, Tensor<T> weights, Tensor<T> offsets, Tensor<T> mask, Tensor<T> bias, int strideH, int strideW, int padH, int padW, int dilationH, int dilationW, int groups, int deformGroups, FusedActivationType activation)
Tensor<T> DeformableConv2DGrouped(Tensor<T> input, Tensor<T> kernel, Tensor<T> offset, Tensor<T> mask, int[] stride, int[] padding, int[] dilation, int groups, int deformGroups)
Tensor<T> DeformableConv2DGroupedBackwardInput(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> kernel, Tensor<T> offset, Tensor<T> mask, int[] inputShape, int[] stride, int[] padding, int[] dilation, int groups, int deformGroups)
Tensor<T> DeformableConv2DGroupedBackwardKernel(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> offset, Tensor<T> mask, int[] kernelShape, int[] stride, int[] padding, int[] dilation, int groups, int deformGroups)
Tensor<T> DeformableConv2DGroupedBackwardMask(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> kernel, Tensor<T> offset, Tensor<T> mask, int[] stride, int[] padding, int[] dilation, int groups, int deformGroups)
Tensor<T> DeformableConv2DGroupedBackwardOffset(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> kernel, Tensor<T> offset, Tensor<T> mask, int[] stride, int[] padding, int[] dilation, int groups, int deformGroups)
Tensor<T> DepthwiseConv1D(Tensor<T> input, Tensor<T> kernel, int stride, int padding)
Tensor<T> DepthwiseConv2D(Tensor<T> input, Tensor<T> kernel, int[] stride, int[] padding)
Tensor<T> DepthwiseConv2DBatchNorm(Tensor<T> input, Tensor<T> kernel, int[] stride, int[] padding, Tensor<T> gamma, Tensor<T> beta, Tensor<T> runningMean, Tensor<T> runningVariance, double epsilon, FusedActivationType activation)
Tensor<T> DepthwiseConv2DGpu(Tensor<T> input, Tensor<T> kernel, Tensor<T> bias, int strideH, int strideW, int padH, int padW, FusedActivationType activation)
Tensor<T> DistanceBoxIou(Tensor<T> boxesA, Tensor<T> boxesB)
Tensor<T> DivideByBroadcastGpu(Tensor<T> a, Tensor<T> b)
Tensor<T> DivideScalarGpu(Tensor<T> input, float scalar)
Tensor<T> Dropout(Tensor<T> input, double dropoutRate, bool training, Tensor<T> mask)
Tensor<T> DropoutBackwardGpu(Tensor<T> gradOutput, Tensor<T> mask, float dropoutRate)
Tensor<T> ELU(Tensor<T> tensor, double alpha)
Tensor<T> EluBackward(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> output, double alpha)
Tensor<T> EluBackwardGpu(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> output, float alpha)
Tensor<T> Embedding(Tensor<int> indices, Tensor<T> embeddingTable)
Tensor<T> EmbeddingBackwardGpu(Tensor<T> gradOutput, Tensor<int> indices, int vocabSize, int embeddingDim)
Tensor<T> EmbeddingLookupGpu(Tensor<T> embeddingTable, Tensor<int> indices)
Tensor<T> EmbeddingLookupGpu(Tensor<T> embeddingTableGpu, Tensor<int> indices, int embeddingDim)
Tensor<T> EmptyInContext(int[] shape, GpuTensorRole role)
Tensor<T> ExpGpu(Tensor<T> input)
Tensor<T> FlashAttention(Tensor<T> query, Tensor<T> key, Tensor<T> value, double? scale, bool isCausal, Tensor<T> softmaxStats, Tensor<T> attentionBias)
Tensor<T> FlashAttentionBackward(Tensor<T> gradOutput, Tensor<T> query, Tensor<T> key, Tensor<T> value, Tensor<T> output, Tensor<T> softmaxStats, double scale, bool isCausal, Tensor<T> gradQuery, Tensor<T> gradKey, Tensor<T> gradValue, Tensor<T> attentionBias)
Tensor<T> Fold(Tensor<T> input, int[] outputSize, int[] kernelSize, int[] stride, int[] padding)
Tensor<T> ForwardSplat(Tensor<T> input, Tensor<T> flow, bool normalize)
Tensor<T> ForwardSplatBackwardFlow(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> flow, Tensor<T> output, bool normalize)
Tensor<T> ForwardSplatBackwardInput(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> flow, bool normalize)
Tensor<T> FusedBatchNorm(Tensor<T> input, Tensor<T> gamma, Tensor<T> beta, Tensor<T> runningMean, Tensor<T> runningVar, double epsilon, double momentum, bool training, FusedActivationType activation, Tensor<T> saveMean, Tensor<T> saveVar)
Tensor<T> FusedBiasDropout(Tensor<T> input, Tensor<T> bias, double dropoutRate, bool training, Tensor<T> mask)
Tensor<T> FusedConv2D(Tensor<T> input, Tensor<T> kernel, Tensor<T> bias, int strideH, int strideW, int padH, int padW, int dilationH, int dilationW, FusedActivationType activation)
Tensor<T> FusedConv2DBatchNorm(Tensor<T> input, Tensor<T> kernel, Tensor<T> bias, Tensor<T> gamma, Tensor<T> beta, Tensor<T> runningMean, Tensor<T> runningVariance, double epsilon, int strideH, int strideW, int padH, int padW, int dilationH, int dilationW, FusedActivationType activation)
Tensor<T> FusedConv2DGpu(Tensor<T> input, Tensor<T> kernel, Tensor<T> bias, int strideH, int strideW, int padH, int padW, int dilationH, int dilationW, FusedActivationType activation)
Tensor<T> FusedConv3D(Tensor<T> input, Tensor<T> kernel, Tensor<T> bias, int strideD, int strideH, int strideW, int padD, int padH, int padW, int dilationD, int dilationH, int dilationW, FusedActivationType activation)
Tensor<T> FusedConv3DGpu(Tensor<T> input, Tensor<T> kernel, Tensor<T> bias, int strideD, int strideH, int strideW, int padD, int padH, int padW, int dilationD, int dilationH, int dilationW, FusedActivationType activation)
Tensor<T> FusedConvTranspose2D(Tensor<T> input, Tensor<T> kernel, Tensor<T> bias, int strideH, int strideW, int padH, int padW, int outputPadH, int outputPadW, FusedActivationType activation)
Tensor<T> FusedConvTranspose2DGpu(Tensor<T> input, Tensor<T> kernel, Tensor<T> bias, int strideH, int strideW, int padH, int padW, int outputPadH, int outputPadW, FusedActivationType activation)
Tensor<T> FusedHierarchicalSoftmax(Tensor<T> input, Tensor<T> nodeWeights, int numClasses)
Tensor<T> FusedLinear(Tensor<T> input, Tensor<T> weights, Tensor<T> bias, FusedActivationType activation, FusedActivationParams activationParams)
Tensor<T> FusedLinearCrossEntropyWithLogits(Tensor<T> hidden, Tensor<T> weight, Tensor<T> bias, Tensor<T> target)
Tensor<T> FusedLinearCrossEntropyWithLogits(Tensor<T> hidden, Tensor<T> weight, Tensor<T> bias, Tensor<int> targetIds)
Tensor<T> FusedLinearCrossEntropyWithLogits(Tensor<T> hidden, Tensor<T> weight, Tensor<T> bias, Tensor<int> targetIds, bool computeReadoutGrads)
Tensor<T> FusedLinearGELU(Tensor<T> input, Tensor<T> weight, Tensor<T> bias)
Tensor<T> FusedLinearGpu(Tensor<T> input, Tensor<T> weights, Tensor<T> bias, FusedActivationType activation)
Tensor<T> FusedLinearMaxout(Tensor<T> input, Tensor<T> weights, Tensor<T> bias, int numPieces)
Tensor<T> FusedLinearReLU(Tensor<T> input, Tensor<T> weight, Tensor<T> bias)
Tensor<T> FusedLinearSigmoid(Tensor<T> input, Tensor<T> weight, Tensor<T> bias)
Tensor<T> FusedLinearSwish(Tensor<T> input, Tensor<T> weight, Tensor<T> bias)
Tensor<T> FusedLinearTanh(Tensor<T> input, Tensor<T> weight, Tensor<T> bias)
Tensor<T> GELU(Tensor<T> tensor)
Tensor<T> GatedDeltaNetScanForward(Tensor<T> qProj, Tensor<T> kProj, Tensor<T> vProj, Tensor<T> alpha, Tensor<T> beta, int numHeads)
Tensor<T> GatherGpu(Tensor<T> source, IGpuBuffer indices, int numIndices, int featureSize)
Tensor<T> GeluBackward(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> GeluBackwardGpu(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> GeneralizedBoxIou(Tensor<T> boxesA, Tensor<T> boxesB)
Tensor<T> GetOrCacheWeightsGpu(Tensor<T> tensor, GpuTensorRole role)
Tensor<T> GlaScanForward(Tensor<T> qProj, Tensor<T> kProj, Tensor<T> vProj, Tensor<T> gate, int numHeads)
Tensor<T> GlobalMaxPoolBackwardGpu(Tensor<T> gradOutput, int[] maxIndices, int[] inputShape)
Tensor<T> GlobalMaxPoolGpu(Tensor<T> input, int[] maxIndices)
Tensor<T> GlobalMaxPoolGpuWithGpuIndices(Tensor<T> input, Tensor<float> gpuIndices)
Tensor<T> GlobalMeanPoolBackwardGpu(Tensor<T> gradOutput, int[] inputShape)
Tensor<T> GlobalMeanPoolGpu(Tensor<T> input)
Tensor<T> GreaterThanScalarGpu(Tensor<T> input, float scalar)
Tensor<T> GridSample(Tensor<T> input, Tensor<T> grid)
Tensor<T> GridSample(Tensor<T> input, Tensor<T> grid, GridSampleMode mode, GridSamplePadding padding, bool alignCorners)
Tensor<T> GridSampleBackwardGrid(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> grid, GridSampleMode mode, GridSamplePadding padding, bool alignCorners)
Tensor<T> GridSampleBackwardInput(Tensor<T> gradOutput, Tensor<T> grid, int[] inputShape, GridSampleMode mode, GridSamplePadding padding, bool alignCorners)
Tensor<T> GridSampleGpu(Tensor<T> input, Tensor<T> grid, int paddingMode, bool alignCorners)
Tensor<T> GroupNorm(Tensor<T> input, int numGroups, Tensor<T> gamma, Tensor<T> beta, double epsilon, Tensor<T> mean, Tensor<T> variance)
Tensor<T> GroupedQueryAttention(Tensor<T> query, Tensor<T> key, Tensor<T> value, int numQueriesPerKV, double? scale, bool isCausal, Tensor<T> attentionWeights)
Tensor<T> GroupedQueryAttentionBackward(Tensor<T> gradOutput, Tensor<T> query, Tensor<T> key, Tensor<T> value, Tensor<T> attentionWeights, int numQueriesPerKV, double scale, Tensor<T> gradQuery, Tensor<T> gradKey, Tensor<T> gradValue)
Tensor<T> GruSequenceForward(Tensor<T> input, Tensor<T> h0, Tensor<T> wIh, Tensor<T> wHh, Tensor<T> bIh, Tensor<T> bHh, Tensor<T> finalHidden, bool returnSequences, bool resetAfter)
Tensor<T> HardSwish(Tensor<T> input)
Tensor<T> HardsigmoidBackward(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> HardswishBackward(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> InstanceNorm(Tensor<T> input, Tensor<T> gamma, Tensor<T> beta, double epsilon, Tensor<T> mean, Tensor<T> variance)
Tensor<T> Interpolate(Tensor<T> input, int[] sizes, InterpolateMode mode, bool alignCorners)
Tensor<T> InterpolateByScale(Tensor<T> input, double[] scaleFactors, InterpolateMode mode, bool alignCorners)
Tensor<T> LayerNorm(Tensor<T> input, Tensor<T> gamma, Tensor<T> beta, double epsilon, Tensor<T> mean, Tensor<T> variance)
Tensor<T> LeakyReLU(Tensor<T> tensor, T alpha)
Tensor<T> LeakyReluBackward(Tensor<T> gradOutput, Tensor<T> input, double negativeSlope)
Tensor<T> LeakyReluBackwardGpu(Tensor<T> gradOutput, Tensor<T> input, float alpha)
Tensor<T> LocallyConnectedConv2D(Tensor<T> input, Tensor<T> weights, Tensor<T> bias, int[] stride)
Tensor<T> LocallyConnectedConv2DBackwardBias(Tensor<T> gradOutput)
Tensor<T> LocallyConnectedConv2DBackwardInput(Tensor<T> gradOutput, Tensor<T> weights, int[] inputShape, int[] stride)
Tensor<T> LocallyConnectedConv2DBackwardWeights(Tensor<T> gradOutput, Tensor<T> input, int[] weightsShape, int[] stride)
Tensor<T> LocallyConnectedConv2DGpu(Tensor<T> input, Tensor<T> weights, Tensor<T> bias, int strideH, int strideW, FusedActivationType activation)
Tensor<T> LstmCellFromGates(Tensor<T> gates, Tensor<T> previousCell, Tensor<T> cell)
Tensor<T> LstmSequenceForward(Tensor<T> input, Tensor<T> h0, Tensor<T> c0, Tensor<T> wIh, Tensor<T> wHh, Tensor<T> bIh, Tensor<T> bHh, Tensor<T> finalHidden, Tensor<T> finalCell, bool returnSequences)
Tensor<T> LstmSequenceForward(Tensor<T> input, Tensor<T> h0, Tensor<T> c0, Tensor<T> wIh, Tensor<T> wHh, Tensor<T> bIh, Tensor<T> bHh, bool returnSequences)
Tensor<T> Mamba2SsdScanForward(Tensor<T> x, Tensor<T> delta, Tensor<T> aLog, Tensor<T> bParam, Tensor<T> cParam, Tensor<T> dParam, int numHeads)
Tensor<T> MambaSelectiveScanForward(Tensor<T> x, Tensor<T> delta, Tensor<T> aLog, Tensor<T> bParam, Tensor<T> cParam, Tensor<T> dParam)
Tensor<T> MatMulGpuTensors(Tensor<T> A, Tensor<T> B)
Tensor<T> MaxAxisGpu(Tensor<T> input, int axis)
Tensor<T> MaxPool2D(Tensor<T> input, int poolSize, int stride, int padding)
Tensor<T> MaxPool2DBackward(Tensor<T> gradOutput, int[] maxIndices, int[] inputShape, int[] poolSize, int[] stride)
Tensor<T> MaxPool2DBackwardGpu(Tensor<T> gradOutput, IGpuBuffer gpuIndices, int[] inputShape, int[] poolSize, int[] stride)
Tensor<T> MaxPool2DBackwardWithTensorIndices(Tensor<T> gradOutput, Tensor<int> maxIndices, int[] inputShape, int[] poolSize, int[] stride)
Tensor<T> MaxPool2DGpu(Tensor<T> input, int[] poolSize, int[] stride, IGpuBuffer gpuIndices)
Tensor<T> MaxPool2DWithIndices(Tensor<T> input, int[] poolSize, int[] stride, int[] maxIndices)
Tensor<T> MaxPool2DWithTensorIndices(Tensor<T> input, int[] poolSize, int[] stride, Tensor<int> maxIndices)
Tensor<T> MaxPool3DBackwardGpu(Tensor<T> gradOutput, IGpuBuffer gpuIndices, int[] inputShape, int[] poolSize, int[] stride)
Tensor<T> MaxPool3DBackwardWithTensorIndices(Tensor<T> gradOutput, Tensor<int> maxIndices, int[] inputShape, int[] poolSize, int[] stride)
Tensor<T> MaxPool3DGpu(Tensor<T> input, int[] poolSize, int[] stride, IGpuBuffer gpuIndices)
Tensor<T> MaxPool3DWithTensorIndices(Tensor<T> input, int[] poolSize, int[] stride, Tensor<int> maxIndices)
Tensor<T> MesaScanForward(Tensor<T> q, Tensor<T> k, Tensor<T> v, Tensor<T> initialWeights, T regularization, int numHeads)
Tensor<T> Mish(Tensor<T> tensor)
Tensor<T> MishBackward(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> MlpForward(Tensor<T> input, IReadOnlyList<Tensor<T>> weights, IReadOnlyList<Tensor<T>> biases, FusedActivationType hiddenActivation, FusedActivationType outputActivation, FusedActivationParams hiddenActivationParams, FusedActivationParams outputActivationParams)
Tensor<T> MuLawDecoding(Tensor<int> input, int quantizationChannels)
Tensor<T> MultiHeadAttentionCore(Tensor<T> query, Tensor<T> key, Tensor<T> value, int numHeads, double? scale, bool causal)
Tensor<T> MultiHeadAttentionForward(Tensor<T> input, Tensor<T> qWeight, Tensor<T> kWeight, Tensor<T> vWeight, Tensor<T> outWeight, int numHeads, Tensor<bool> mask)
Tensor<T> MultiplyGpu(Tensor<T> a, Tensor<T> b)
Tensor<T> MultiresolutionHashEncoding(Tensor<T> positions, Tensor<T>[] hashTables, int baseResolution, double perLevelScale, int featuresPerLevel, HashGridAlignment alignment)
Tensor<T> NativeAtan2(Tensor<T> imag, Tensor<T> real)
Tensor<T> NativeBatchedCavityForward(Tensor<T> input, Tensor<Complex<T>> cavityFilters, int numBounces)
Tensor<T> NativeComplexIFFT2DReal(Tensor<Complex<T>> input)
Tensor<T> NativeComplexIFFTNDReal(Tensor<Complex<T>> input, int[] axes)
Tensor<T> NativeComplexIFFTReal(Tensor<Complex<T>> input)
Tensor<T> NativeComplexMagnitude(Tensor<Complex<T>> a)
Tensor<T> NativeComplexMagnitudeSquared(Tensor<Complex<T>> a)
Tensor<T> NativeComplexPhase(Tensor<Complex<T>> a)
Tensor<T> NativeExp(Tensor<T> input)
Tensor<T> NativeMagnitudeAndPhase(Tensor<Complex<T>> input, Tensor<T> phase)
Tensor<T> NativeMfccFeatures(Tensor<T> waveforms, int numSegments, int numMfcc, int paddedDim)
Tensor<T> NativeNormalizeRows(Tensor<T> input, bool inPlace)
Tensor<T> NativePacFeatures(Tensor<T> waveforms, int sampleRate, int envelopeRate, double thetaLow, double thetaHigh, ValueTuple<double, double>[] gammaBands)
Tensor<T> NativeSpectralFilter(Tensor<T> input, Tensor<Complex<T>> filter)
Tensor<T> NativeSpectralFilterBatch(Tensor<T> input, Tensor<Complex<T>> filter)
Tensor<T> NativeTanh(Tensor<T> input)
Tensor<T> NativeWidebandFeatures(Tensor<T> waveforms, int numSegments, int numBins)
Tensor<T> NearestNeighborUpsample3DBackwardGpu(Tensor<T> gradOutput, int[] inputShape, int scaleDepth, int scaleHeight, int scaleWidth)
Tensor<T> NearestNeighborUpsample3DGpu(Tensor<T> input, int scaleDepth, int scaleHeight, int scaleWidth)
Tensor<T> OctonionMatMulTensor(Tensor<T> input, Tensor<T> weight)
Tensor<T> PadNd(Tensor<T> input, int[] pad, PadMode mode, T value)
Tensor<T> PartialCorrelationVolume(Tensor<T> first, Tensor<T> second, int radius)
Tensor<T> PermuteGpu(Tensor<T> input, int[] permutation)
Tensor<T> PermuteResidentGpu(Tensor<T> tensor, int[] axes)
Tensor<T> PitchShift(Tensor<T> waveform, int sampleRate, double nSteps, int nFft, int hopLength)
Tensor<T> PsRoIAlign(Tensor<T> input, Tensor<T> boxes, int outputHeight, int outputWidth, int outputChannels, float spatialScale, int samplingRatio)
Tensor<T> PsRoIPool(Tensor<T> input, Tensor<T> boxes, int outputHeight, int outputWidth, int outputChannels, float spatialScale)
Tensor<T> RMSNorm(Tensor<T> input, Tensor<T> gamma, double epsilon, Tensor<T> rms)
Tensor<T> RandomNormalGpu(int[] shape, float mean, float stdDev, ulong seed)
Tensor<T> RandomUniformGpu(int[] shape, float min, float max, ulong seed)
Tensor<T> RbfKernelGpu(Tensor<T> input, Tensor<T> centers, Tensor<T> widths)
Tensor<T> ReLU(Tensor<T> tensor)
Tensor<T> ReciprocalBackward(Tensor<T> gradOutput, Tensor<T> output)
Tensor<T> ReduceMax(Tensor<T> input, int[] axes, bool keepDims)
Tensor<T> ReduceMax(Tensor<T> input, int[] axes, bool keepDims, int[] maxIndices)
Tensor<T> ReduceMaxBackwardWithTensorIndices(Tensor<T> gradOutput, Tensor<int> maxIndices, int[] inputShape, int[] axes)
Tensor<T> ReduceMaxWithTensorIndices(Tensor<T> input, int[] axes, bool keepDims, Tensor<int> maxIndices)
Tensor<T> ReduceMean(Tensor<T> input, int[] axes, bool keepDims)
Tensor<T> ReduceStd(Tensor<T> input, int[] axes, bool keepDims)
Tensor<T> ReduceSum(Tensor<T> tensor, int[] axes, bool keepDims)
Tensor<T> Relu6Backward(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> ReluBackward(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> ReluBackwardGpu(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> ReluGpu(Tensor<T> input)
Tensor<T> ReorderToNchw(Tensor<T> tensor)
Tensor<T> ReorderToNchwc(Tensor<T> tensor, TensorLayout targetLayout)
Tensor<T> Resample(Tensor<T> waveform, int origRate, int newRate)
Tensor<T> Reshape(Tensor<T> tensor, int[] newShape)
Tensor<T> ReshapeGpu(Tensor<T> input, int[] newShape)
Tensor<T> RgLruScanForward(Tensor<T> value, Tensor<T> recGate, Tensor<T> inpGate, Tensor<T> decay)
Tensor<T> RoIAlign(Tensor<T> input, Tensor<T> boxes, int outputHeight, int outputWidth, float spatialScale, int samplingRatio, bool aligned)
Tensor<T> RoIPool(Tensor<T> input, Tensor<T> boxes, int outputHeight, int outputWidth, float spatialScale)
Tensor<T> RoutedDiagonalSsmScanForward(Tensor<T> input, Tensor<T> activeMask, Tensor<T> transition, Tensor<T> inputMap, Tensor<T> outputMap, Tensor<T> skip)
Tensor<T> Rwkv4WkvForward(Tensor<T> rProj, Tensor<T> kProj, Tensor<T> vProj, Tensor<T> timeDecay, Tensor<T> timeFirst)
Tensor<T> Rwkv7SequenceForward(Tensor<T> rProj, Tensor<T> kappa, Tensor<T> kTilde, Tensor<T> vProj, Tensor<T> decayLogit, Tensor<T> iclRate, int numHeads)
Tensor<T> ScaleGpu(Tensor<T> input, float scalar)
Tensor<T> ScaledDotProductAttention(Tensor<T> query, Tensor<T> key, Tensor<T> value, Tensor<bool> mask, double? scale, bool isCausal, double softcap)
Tensor<T> ScaledDotProductAttentionGpu(Tensor<T> query, Tensor<T> key, Tensor<T> value, double scale, Tensor<T> attentionWeights, bool isCausal)
Tensor<T> ScaledDotProductAttentionGpu(Tensor<T> query, Tensor<T> key, Tensor<T> value, double scale, bool isCausal, float softcap)
Tensor<T> ScatterAddGpu(Tensor<T> nodeFeatures, int[] sourceIndices, int[] targetIndices, float[] edgeValues)
Tensor<T> ScatterAddGpu(Tensor<T> source, IGpuBuffer indices, int destSize)
Tensor<T> ScatterMean(Tensor<T> source, Tensor<int> indices, Tensor<int> counts, int dim, int? outputSize)
Tensor<T> SeluBackward(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> Sigmoid(Tensor<T> tensor)
Tensor<T> SigmoidBackward(Tensor<T> gradOutput, Tensor<T> output)
Tensor<T> SigmoidBackwardGpu(Tensor<T> gradOutput, Tensor<T> output)
Tensor<T> SinGpu(Tensor<T> input)
Tensor<T> SliceColumnGpu(Tensor<T> input, int columnIndex)
Tensor<T> SliceGpu(Tensor<T> input, int axis, int start, int end)
Tensor<T> Softmax(Tensor<T> input, int axis)
Tensor<T> SoftmaxBackwardGpu(Tensor<T> gradOutput, Tensor<T> output)
Tensor<T> SoftmaxGpu(Tensor<T> input)
Tensor<T> Softplus(Tensor<T> input)
Tensor<T> SoftplusBackward(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> SparseDenseMatMulBiasGpu(SparseTensor<T> sparseA, Tensor<T> denseB, Tensor<T> bias)
Tensor<T> SparseDenseMatMulGpu(SparseTensor<T> sparseA, Tensor<T> denseB)
Tensor<T> Spectrogram(Tensor<T> waveform, int nFft, int hopLength, int winLength, Tensor<T> window)
Tensor<T> SquashBackwardGpu(Tensor<T> gradOutput, Tensor<T> input, int axis, float epsilon)
Tensor<T> SquashGpu(Tensor<T> input, int axis, float epsilon)
Tensor<T> StdBackward(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> mean, Tensor<T> std, int[] axes)
Tensor<T> StftPhase(Tensor<T> waveform, int nFft, int hopLength, int winLength, Tensor<T> window)
Tensor<T> StopGradient(Tensor<T> tensor)
Tensor<T> SubtractGpu(Tensor<T> a, Tensor<T> b)
Tensor<T> SumAxisGpu(Tensor<T> input, int axis)
Tensor<T> Swish(Tensor<T> tensor)
Tensor<T> SwishBackward(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> SwishBackwardGpu(Tensor<T> gradOutput, Tensor<T> input)
Tensor<T> Tanh(Tensor<T> tensor)
Tensor<T> TanhBackward(Tensor<T> gradOutput, Tensor<T> output)
Tensor<T> TanhBackwardGpu(Tensor<T> gradOutput, Tensor<T> output)
Tensor<T> TanhGpu(Tensor<T> input)
Tensor<T> TensorAbs(Tensor<T> tensor)
Tensor<T> TensorAcos(Tensor<T> tensor)
Tensor<T> TensorAcosh(Tensor<T> tensor)
Tensor<T> TensorAdaptiveAvgPool1D(Tensor<T> input, int outputSize)
Tensor<T> TensorAdaptiveAvgPool3D(Tensor<T> input, int[] outputSize)
Tensor<T> TensorAdaptiveMaxPool1D(Tensor<T> input, int outputSize)
Tensor<T> TensorAdaptiveMaxPool2D(Tensor<T> input, int[] outputSize)
Tensor<T> TensorAdaptiveMaxPool3D(Tensor<T> input, int[] outputSize)
Tensor<T> TensorAdd(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorAddActivation(Tensor<T> a, Tensor<T> b, FusedActivationType activation)
Tensor<T> TensorAddMM(Tensor<T> input, Tensor<T> a, Tensor<T> b, T alpha, T beta)
Tensor<T> TensorAddScalar(Tensor<T> tensor, T scalar)
Tensor<T> TensorAddScaled(Tensor<T> a, Tensor<T> b, T scaleA, T scaleB)
Tensor<T> TensorAddbmm(Tensor<T> input, Tensor<T> batch1, Tensor<T> batch2, double beta, double alpha)
Tensor<T> TensorAddcdiv(Tensor<T> input, Tensor<T> tensor1, Tensor<T> tensor2, double value)
Tensor<T> TensorAddcmul(Tensor<T> input, Tensor<T> tensor1, Tensor<T> tensor2, double value)
Tensor<T> TensorAddmv(Tensor<T> input, Tensor<T> mat, Tensor<T> vec, double beta, double alpha)
Tensor<T> TensorAddr(Tensor<T> input, Tensor<T> vec1, Tensor<T> vec2, double beta, double alpha)
Tensor<T> TensorAiryAi(Tensor<T> tensor)
Tensor<T> TensorAll(Tensor<T> tensor, int[] axes, bool keepDims)
Tensor<T> TensorAlphaDropout(Tensor<T> tensor, double p, bool training, int? seed)
Tensor<T> TensorAmin(Tensor<T> tensor, int[] axes, bool keepDims)
Tensor<T> TensorAngle(Tensor<T> input)
Tensor<T> TensorAny(Tensor<T> tensor, int[] axes, bool keepDims)
Tensor<T> TensorArange(double start, double end, double step)
Tensor<T> TensorAsin(Tensor<T> tensor)
Tensor<T> TensorAsinh(Tensor<T> tensor)
Tensor<T> TensorAtLeast1D(Tensor<T> tensor)
Tensor<T> TensorAtLeast2D(Tensor<T> tensor)
Tensor<T> TensorAtLeast3D(Tensor<T> tensor)
Tensor<T> TensorAtan(Tensor<T> tensor)
Tensor<T> TensorAtan2(Tensor<T> y, Tensor<T> x)
Tensor<T> TensorAtanh(Tensor<T> tensor)
Tensor<T> TensorAvgPool1D(Tensor<T> input, int kernelSize, int stride)
Tensor<T> TensorAvgPool2D(Tensor<T> input, int poolSize, int stride, int padding, bool countIncludePad)
Tensor<T> TensorBCEWithLogitsLoss(Tensor<T> logits, Tensor<T> targets)
Tensor<T> TensorBaddbmm(Tensor<T> input, Tensor<T> batch1, Tensor<T> batch2, double beta, double alpha)
Tensor<T> TensorBartlettWindow(int length, bool periodic)
Tensor<T> TensorBentIdentity(Tensor<T> tensor)
Tensor<T> TensorBernoulli(Tensor<T> probabilities, int? seed)
Tensor<T> TensorBernoulliFromRng(Random generator, Tensor<T> probabilities, UniformDrawComparison successWhen)
Tensor<T> TensorBesselJ0(Tensor<T> tensor)
Tensor<T> TensorBesselJ1(Tensor<T> tensor)
Tensor<T> TensorBesselY0(Tensor<T> tensor)
Tensor<T> TensorBesselY1(Tensor<T> tensor)
Tensor<T> TensorBilinear(Tensor<T> input1, Tensor<T> input2, Tensor<T> weight, Tensor<T> bias)
Tensor<T> TensorBinomial(Tensor<T> count, Tensor<T> probability, int? seed)
Tensor<T> TensorBitwiseAnd(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorBitwiseLeftShift(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorBitwiseNot(Tensor<T> tensor)
Tensor<T> TensorBitwiseOr(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorBitwiseRightShift(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorBitwiseXor(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorBlackmanWindow(int length, bool periodic)
Tensor<T> TensorBlockDiag(Tensor<T>[] matrices)
Tensor<T> TensorBroadcastAdd(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorBroadcastDivide(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorBroadcastMultiply(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorBroadcastSubtract(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorBroadcastTo(Tensor<T> input, int[] targetShape)
Tensor<T> TensorCDist(Tensor<T> x1, Tensor<T> x2, double p)
Tensor<T> TensorCIoULoss(Tensor<T> predicted, Tensor<T> target)
Tensor<T> TensorCTCLabelPosteriors(Tensor<T> logProbs, Tensor<int> targets, int[] inputLengths, int[] targetLengths, int blank)
Tensor<T> TensorCartesianProd(Tensor<T>[] tensors)
Tensor<T> TensorCauchy(int[] shape, double median, double sigma, int? seed)
Tensor<T> TensorCeiling(Tensor<T> tensor)
Tensor<T> TensorCelu(Tensor<T> tensor, double alpha)
Tensor<T> TensorChannelBiasAdd(Tensor<T> input, Tensor<T> bias)
Tensor<T> TensorChannelDropout(Tensor<T> tensor, double p, bool training, int spatialDims, int? seed)
Tensor<T> TensorChannelShuffle(Tensor<T> tensor, int groups)
Tensor<T> TensorChebyshevPolynomialT(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorChebyshevPolynomialU(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorChebyshevPolynomialV(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorChebyshevPolynomialW(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorClamp(Tensor<T> tensor, T min, T max)
Tensor<T> TensorClampMax(Tensor<T> tensor, T max)
Tensor<T> TensorClampMin(Tensor<T> tensor, T min)
Tensor<T> TensorClampTensor(Tensor<T> tensor, Tensor<T> min, Tensor<T> max)
Tensor<T> TensorColumnStack(Tensor<T>[] tensors)
Tensor<T> TensorCombinations(Tensor<T> tensor, int r, bool withReplacement)
Tensor<T> TensorConcatenate(Tensor<T>[] tensors, int axis)
Tensor<T> TensorConstantPad(Tensor<T> tensor, int[] padding, T value)
Tensor<T> TensorConv2D(Tensor<T> input, Tensor<T> kernel, int stride, int padding, int dilation)
Tensor<T> TensorConvTbc(Tensor<T> input, Tensor<T> weight, Tensor<T> bias, int pad)
Tensor<T> TensorConvTranspose1D(Tensor<T> input, Tensor<T> kernel, int stride, int padding, int outputPadding)
Tensor<T> TensorConvolution(Tensor<T> input, Tensor<T> weight, Tensor<T> bias, int[] stride, int[] padding, int[] dilation, bool transposed, int[] outputPadding, int groups)
Tensor<T> TensorCopysign(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorCorrcoef(Tensor<T> tensor)
Tensor<T> TensorCos(Tensor<T> tensor)
Tensor<T> TensorCosineSimilarity(Tensor<T> x1, Tensor<T> x2, int dim, double eps)
Tensor<T> TensorCountIndices(Tensor<int> indices, int numBins)
Tensor<T> TensorCov(Tensor<T> tensor, int correction)
Tensor<T> TensorCrfForward(Tensor<T> emissions, Tensor<T> transitions, Tensor<T> startScores, Tensor<T> endScores)
Tensor<T> TensorCross(Tensor<T> a, Tensor<T> b, int dim)
Tensor<T> TensorCrossEntropyLoss(Tensor<T> logits, Tensor<T> targets)
Tensor<T> TensorCumMax(Tensor<T> tensor, int axis)
Tensor<T> TensorCumMin(Tensor<T> tensor, int axis)
Tensor<T> TensorCumProd(Tensor<T> tensor, int axis)
Tensor<T> TensorCumSum(Tensor<T> tensor, int axis)
Tensor<T> TensorCumulativeTrapezoid(Tensor<T> y, double dx, int dim)
Tensor<T> TensorDIoULoss(Tensor<T> predicted, Tensor<T> target)
Tensor<T> TensorDStack(Tensor<T>[] tensors)
Tensor<T> TensorDeg2Rad(Tensor<T> tensor)
Tensor<T> TensorDiagEmbed(Tensor<T> tensor, int offset)
Tensor<T> TensorDiagflat(Tensor<T> tensor, int offset)
Tensor<T> TensorDiagonalScatter(Tensor<T> input, Tensor<T> src, int offset, int dim1, int dim2)
Tensor<T> TensorDiff(Tensor<T> tensor, int n, int dim)
Tensor<T> TensorDigamma(Tensor<T> tensor)
Tensor<T> TensorDivide(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorDivideScalar(Tensor<T> tensor, T scalar)
Tensor<T> TensorDot(Tensor<T> a, Tensor<T> b, int[] axesA, int[] axesB)
Tensor<T> TensorEmptyLike(Tensor<T> tensor)
Tensor<T> TensorEntr(Tensor<T> tensor)
Tensor<T> TensorErf(Tensor<T> tensor)
Tensor<T> TensorErfc(Tensor<T> tensor)
Tensor<T> TensorErfcx(Tensor<T> tensor)
Tensor<T> TensorErfinv(Tensor<T> tensor)
Tensor<T> TensorExp(Tensor<T> tensor)
Tensor<T> TensorExp2(Tensor<T> tensor)
Tensor<T> TensorExpandAs(Tensor<T> tensor, Tensor<T> other)
Tensor<T> TensorExpandDims(Tensor<T> tensor, int axis)
Tensor<T> TensorExpm1(Tensor<T> tensor)
Tensor<T> TensorExponential(int[] shape, double rate, int? seed)
Tensor<T> TensorFakeQuantize(Tensor<T> tensor, int numBits, double? scale, int? zeroPoint, bool symmetric)
Tensor<T> TensorFeatureAlphaDropout(Tensor<T> tensor, double p, bool training, int? seed)
Tensor<T> TensorFlatten(Tensor<T> tensor)
Tensor<T> TensorFlip(Tensor<T> tensor, int[] axes)
Tensor<T> TensorFliplr(Tensor<T> tensor)
Tensor<T> TensorFlipud(Tensor<T> tensor)
Tensor<T> TensorFloatPower(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorFloor(Tensor<T> tensor)
Tensor<T> TensorFloorDivide(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorFmax(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorFmin(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorFmod(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorFritschCarlsonLimit(Tensor<T> slopes, Tensor<T> secants)
Tensor<T> TensorFromBytes(ReadOnlySpan<byte> data, NumericStorageType sourceType, int[] shape, bool littleEndian)
Tensor<T> TensorFullLike(Tensor<T> tensor, T value)
Tensor<T> TensorGELU(Tensor<T> tensor)
Tensor<T> TensorGIoULoss(Tensor<T> predicted, Tensor<T> target)
Tensor<T> TensorGammaFromRng(Random generator, Tensor<T> shape, Tensor<T> scale, BoxMullerVariant normal)
Tensor<T> TensorGammaFromRng(Random generator, double shape, Tensor<T> scale, BoxMullerVariant normal)
Tensor<T> TensorGather(Tensor<T> source, Tensor<int> indices, int axis)
Tensor<T> TensorGatherClassValues(Tensor<T> values, Tensor<T> classIndices)
Tensor<T> TensorGatherRowValues(Tensor<T> values, Tensor<int> indices)
Tensor<T> TensorGatherRowValuesBackward(Tensor<T> gradOutput, Tensor<int> indices, int numCols)
Tensor<T> TensorGatherRows(Tensor<T> table, Tensor<long> rows)
Tensor<T> TensorGaussian(Tensor<T> tensor)
Tensor<T> TensorGcd(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorGeometric(int[] shape, double p, int? seed)
Tensor<T> TensorGreaterEqual(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorGridSample3D(Tensor<T> input, Tensor<T> grid, GridSampleMode mode, GridSamplePadding padding, bool alignCorners)
Tensor<T> TensorGroupedMatMul(Tensor<T> a, Tensor<T> b, Tensor<int> offsets, Tensor<T> bias)
Tensor<T> TensorGruCell(Tensor<T> input, Tensor<T> hidden, Tensor<T> wIh, Tensor<T> wHh, Tensor<T> bIh, Tensor<T> bHh)
Tensor<T> TensorHStack(Tensor<T>[] tensors)
Tensor<T> TensorHammingWindow(int length, bool periodic, double alpha, double beta)
Tensor<T> TensorHannWindow(int length, bool periodic)
Tensor<T> TensorHardSigmoid(Tensor<T> tensor)
Tensor<T> TensorHardSwish(Tensor<T> tensor)
Tensor<T> TensorHardtanh(Tensor<T> tensor, double minValue, double maxValue)
Tensor<T> TensorHeaviside(Tensor<T> tensor, Tensor<T> values)
Tensor<T> TensorHermitePolynomialH(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorHermitePolynomialHe(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorHistc(Tensor<T> input, int bins, T min, T max)
Tensor<T> TensorHuberLoss(Tensor<T> predictions, Tensor<T> targets, double delta)
Tensor<T> TensorHyperbolicLinear(Tensor<T> input, Tensor<T> weights, Tensor<T> bias, double curvature)
Tensor<T> TensorHypot(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorI0(Tensor<T> tensor)
Tensor<T> TensorI0e(Tensor<T> tensor)
Tensor<T> TensorI1(Tensor<T> tensor)
Tensor<T> TensorI1e(Tensor<T> tensor)
Tensor<T> TensorIgamma(Tensor<T> a, Tensor<T> x)
Tensor<T> TensorIgammac(Tensor<T> a, Tensor<T> x)
Tensor<T> TensorImag(Tensor<Complex<T>> input)
Tensor<T> TensorIndex(Tensor<T> tensor, Tensor<bool> mask)
Tensor<T> TensorIndex(Tensor<T> tensor, Tensor<int>[] indices)
Tensor<T> TensorIndexAdd(Tensor<T> tensor, int axis, Tensor<int> indices, Tensor<T> source)
Tensor<T> TensorIndexCopy(Tensor<T> tensor, int axis, Tensor<int> indices, Tensor<T> source)
Tensor<T> TensorIndexFill(Tensor<T> tensor, int axis, Tensor<int> indices, T value)
Tensor<T> TensorIndexPut(Tensor<T> tensor, Tensor<int>[] indices, Tensor<T> source, bool accumulate)
Tensor<T> TensorIndexReduce(Tensor<T> tensor, int dim, Tensor<int> index, Tensor<T> source, ScatterReduceMode reduce, bool includeSelf)
Tensor<T> TensorIndexSelect(Tensor<T> tensor, Tensor<int> indices, int axis)
Tensor<T> TensorInner(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorIoULoss(Tensor<T> predicted, Tensor<T> target)
Tensor<T> TensorIsNegInf(Tensor<T> tensor)
Tensor<T> TensorIsPosInf(Tensor<T> tensor)
Tensor<T> TensorIsReal(Tensor<T> tensor)
Tensor<T> TensorIsru(Tensor<T> tensor, double alpha)
Tensor<T> TensorKLDivLoss(Tensor<T> input, Tensor<T> target)
Tensor<T> TensorKaiserWindow(int length, bool periodic, double beta)
Tensor<T> TensorKron(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorL1Loss(Tensor<T> predictions, Tensor<T> targets)
Tensor<T> TensorLaguerrePolynomialL(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorLayerNorm(Tensor<T> input, Tensor<T> gamma, Tensor<T> beta, double epsilon)
Tensor<T> TensorLcm(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorLdexp(Tensor<T> x, Tensor<int> exp)
Tensor<T> TensorLeakyReLU(Tensor<T> tensor, T alpha)
Tensor<T> TensorLegendrePolynomialP(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorLerp(Tensor<T> a, Tensor<T> b, T t)
Tensor<T> TensorLessEqual(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorLgamma(Tensor<T> tensor)
Tensor<T> TensorLinearFilter(Tensor<T> b, Tensor<T> a, Tensor<T> x, Tensor<T> zi)
Tensor<T> TensorLinearFilterGuarded(Tensor<T> b, Tensor<T> a, Tensor<T> x, Tensor<T> zi, double maxAbs)
Tensor<T> TensorLisht(Tensor<T> tensor)
Tensor<T> TensorLocalResponseNorm(Tensor<T> tensor, int size, double alpha, double beta, double k)
Tensor<T> TensorLog(Tensor<T> tensor)
Tensor<T> TensorLog10(Tensor<T> tensor)
Tensor<T> TensorLog1p(Tensor<T> tensor)
Tensor<T> TensorLog2(Tensor<T> tensor)
Tensor<T> TensorLogAddExp(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorLogAddExp2(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorLogCumSumExp(Tensor<T> tensor, int axis)
Tensor<T> TensorLogNdtr(Tensor<T> tensor)
Tensor<T> TensorLogNormal(int[] shape, double mean, double std, int? seed)
Tensor<T> TensorLogSigmoid(Tensor<T> tensor)
Tensor<T> TensorLogSoftmax(Tensor<T> tensor, int axis)
Tensor<T> TensorLogSoftmin(Tensor<T> tensor, int axis)
Tensor<T> TensorLogSumExp(Tensor<T> tensor, int axis, bool keepDims)
Tensor<T> TensorLogit(Tensor<T> tensor, double? eps)
Tensor<T> TensorLogspace(double start, double end, int steps, double logBase)
Tensor<T> TensorLpPool(Tensor<T> input, double power, int[] kernelSize, int[] stride)
Tensor<T> TensorMSELoss(Tensor<T> predictions, Tensor<T> targets)
Tensor<T> TensorMaskedFill(Tensor<T> tensor, Tensor<Bit> mask, T value)
Tensor<T> TensorMaskedFill(Tensor<T> tensor, Tensor<bool> mask, T value)
Tensor<T> TensorMaskedScatter(Tensor<T> tensor, Tensor<Bit> mask, Tensor<T> source)
Tensor<T> TensorMaskedSelect(Tensor<T> tensor, Tensor<Bit> mask)
Tensor<T> TensorMatMul(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorMatMul2DWithPrePackedB(Tensor<T> a, Tensor<T> b, WeightPackHandle prePackedB)
Tensor<T> TensorMatMulTransposed(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorMax(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorMaxPool1D(Tensor<T> input, int kernelSize, int stride)
Tensor<T> TensorMaxPool2D(Tensor<T> input, int poolSize, int stride, int padding)
Tensor<T> TensorMaxUnpool(Tensor<T> input, Tensor<int> indices, int[] outputSize)
Tensor<T> TensorMin(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorMish(Tensor<T> tensor)
Tensor<T> TensorModifiedBesselI0(Tensor<T> tensor)
Tensor<T> TensorModifiedBesselI1(Tensor<T> tensor)
Tensor<T> TensorModifiedBesselK0(Tensor<T> tensor)
Tensor<T> TensorModifiedBesselK1(Tensor<T> tensor)
Tensor<T> TensorMoveDim(Tensor<T> tensor, int source, int destination)
Tensor<T> TensorMsort(Tensor<T> tensor)
Tensor<T> TensorMultiDot(Tensor<T>[] matrices)
Tensor<T> TensorMultinomial(Tensor<T> probabilities, int numSamples, bool replacement, int? seed)
Tensor<T> TensorMultiply(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorMultiplyScalar(Tensor<T> tensor, T scalar)
Tensor<T> TensorMvlgamma(Tensor<T> tensor, int p)
Tensor<T> TensorNLLLoss(Tensor<T> logProbs, Tensor<T> targets)
Tensor<T> TensorNanMean(Tensor<T> tensor, int[] axes, bool keepDims)
Tensor<T> TensorNanQuantile(Tensor<T> tensor, double q, int? dim, bool keepDim, QuantileInterpolation interpolation)
Tensor<T> TensorNanSum(Tensor<T> tensor, int[] axes, bool keepDims)
Tensor<T> TensorNanToNum(Tensor<T> tensor, double? nan, double? posinf, double? neginf)
Tensor<T> TensorNarrow(Tensor<T> tensor, int dim, int start, int length)
Tensor<T> TensorNdtr(Tensor<T> tensor)
Tensor<T> TensorNdtri(Tensor<T> tensor)
Tensor<T> TensorNegate(Tensor<T> tensor)
Tensor<T> TensorNextAfter(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorNormExceptDim(Tensor<T> tensor, double pow, int dim)
Tensor<T> TensorNormal(Tensor<T> mean, Tensor<T> std, int? seed)
Tensor<T> TensorNormalFromRng(Random generator, int[] shape, Func<Random, double> sampler)
Tensor<T> TensorOnesLike(Tensor<T> tensor)
Tensor<T> TensorPDist(Tensor<T> input, double p)
Tensor<T> TensorPReLU(Tensor<T> tensor, Tensor<T> alpha)
Tensor<T> TensorPadSequence(Tensor<T>[] sequences, bool batchFirst, double paddingValue)
Tensor<T> TensorParse(IReadOnlyList<string> tokens, int[] shape, NumberStyles styles, IFormatProvider provider, T missingValue)
Tensor<T> TensorParse(ReadOnlySpan<byte> utf8Text, byte delimiter, int[] shape, NumberStyles styles, IFormatProvider provider, T missingValue)
Tensor<T> TensorPermute(Tensor<T> tensor, int[] axes)
Tensor<T> TensorPixelUnshuffle(Tensor<T> tensor, int downscaleFactor)
Tensor<T> TensorPoisson(Tensor<T> rates, int? seed)
Tensor<T> TensorPolygamma(int n, Tensor<T> tensor)
Tensor<T> TensorPositive(Tensor<T> tensor)
Tensor<T> TensorPower(Tensor<T> tensor, T exponent)
Tensor<T> TensorPut(Tensor<T> tensor, Tensor<int> indices, Tensor<T> source)
Tensor<T> TensorQuantile(Tensor<T> tensor, double q, int? dim, bool keepDim, QuantileInterpolation interpolation)
Tensor<T> TensorRReLU(Tensor<T> tensor, double lower, double upper, bool training, int? seed)
Tensor<T> TensorRad2Deg(Tensor<T> tensor)
Tensor<T> TensorRandLike(Tensor<T> tensor, int? seed)
Tensor<T> TensorRandint(long low, long high, int[] shape, int? seed)
Tensor<T> TensorRandintLike(Tensor<T> tensor, long low, long high, int? seed)
Tensor<T> TensorRandnLike(Tensor<T> tensor, int? seed)
Tensor<T> TensorRandomUniformFromGenerator(Random generator, int[] shape, double min, double max)
Tensor<T> TensorRandperm(int n, int? seed)
Tensor<T> TensorRange(double start, double end, double step)
Tensor<T> TensorReLU(Tensor<T> tensor)
Tensor<T> TensorReLU6(Tensor<T> tensor)
Tensor<T> TensorReal(Tensor<Complex<T>> input)
Tensor<T> TensorRemainder(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorRenorm(Tensor<T> tensor, double p, int dim, double maxNorm)
Tensor<T> TensorRepeatInterleave(Tensor<T> tensor, int repeats, int dim)
Tensor<T> TensorRnnCell(Tensor<T> input, Tensor<T> hidden, Tensor<T> wIh, Tensor<T> wHh, Tensor<T> bIh, Tensor<T> bHh, RnnCellType cell)
Tensor<T> TensorRoll(Tensor<T> tensor, int[] shifts, int[] axes)
Tensor<T> TensorRot90(Tensor<T> tensor, int k, int[] axes)
Tensor<T> TensorRound(Tensor<T> tensor)
Tensor<T> TensorRowStack(Tensor<T>[] tensors)
Tensor<T> TensorRrelu(Tensor<T> tensor, double lower, double upper, bool training, int? seed)
Tensor<T> TensorRsqrt(Tensor<T> tensor)
Tensor<T> TensorRsub(Tensor<T> input, Tensor<T> other, double alpha)
Tensor<T> TensorSELU(Tensor<T> tensor)
Tensor<T> TensorScaledGroupedMatMul(Tensor<T> a, Tensor<T> b, Tensor<T> scaleA, ScalingRecipe recipeA, Tensor<T> scaleB, ScalingRecipe recipeB, ScaleSwizzle swizzleA, ScaleSwizzle swizzleB, Tensor<T> bias, Tensor<int> offsets)
Tensor<T> TensorScaledMatMul(Tensor<T> a, Tensor<T> b, Tensor<T> scaleA, ScalingRecipe recipeA, Tensor<T> scaleB, ScalingRecipe recipeB, ScaleSwizzle swizzleA, ScaleSwizzle swizzleB, Tensor<T> bias)
Tensor<T> TensorScaledModifiedBesselK0(Tensor<T> tensor)
Tensor<T> TensorScaledModifiedBesselK1(Tensor<T> tensor)
Tensor<T> TensorScaledTanh(Tensor<T> tensor, double alpha, double beta)
Tensor<T> TensorScatterAdd(Tensor<T> destination, Tensor<int> indices, Tensor<T> updates, int axis)
Tensor<T> TensorScatterReduce(Tensor<T> tensor, int dim, Tensor<int> indices, Tensor<T> source, ScatterReduceMode mode, bool includeSelf)
Tensor<T> TensorSelect(Tensor<T> tensor, int dim, int index)
Tensor<T> TensorSelectScatter(Tensor<T> tensor, Tensor<T> source, int dim, int index)
Tensor<T> TensorShiftedChebyshevPolynomialT(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorShiftedChebyshevPolynomialU(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorShiftedChebyshevPolynomialV(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorShiftedChebyshevPolynomialW(Tensor<T> x, Tensor<T> n)
Tensor<T> TensorSiLU(Tensor<T> tensor)
Tensor<T> TensorSigmoid(Tensor<T> tensor)
Tensor<T> TensorSign(Tensor<T> tensor)
Tensor<T> TensorSignbit(Tensor<T> tensor)
Tensor<T> TensorSin(Tensor<T> tensor)
Tensor<T> TensorSinc(Tensor<T> tensor)
Tensor<T> TensorSinusoidalTimeEmbedding(Tensor<T> timesteps, int embeddingDim)
Tensor<T> TensorSlice(Tensor<T> tensor, int[] start, int[] length)
Tensor<T> TensorSliceScatter(Tensor<T> tensor, Tensor<T> source, int dim, int start, int length)
Tensor<T> TensorSoftKnn(Tensor<T> input, Tensor<T> supportVectors, Tensor<T> labels, double temperature)
Tensor<T> TensorSoftLocallyWeighted(Tensor<T> input, Tensor<T> xTrain, Tensor<T> yTrain, double bandwidth)
Tensor<T> TensorSoftMarginLoss(Tensor<T> input, Tensor<T> target, LossReduction reduction)
Tensor<T> TensorSoftSplit(Tensor<T> input, Tensor<T> leftValue, Tensor<T> rightValue, int featureIndex, double threshold, double temperature)
Tensor<T> TensorSoftmaxCrossEntropy(Tensor<T> logits, Tensor<T> target)
Tensor<T> TensorSoftmaxRows(Tensor<T> input)
Tensor<T> TensorSoftmin(Tensor<T> tensor, int axis)
Tensor<T> TensorSoftsign(Tensor<T> tensor)
Tensor<T> TensorSphericalBesselJ0(Tensor<T> tensor)
Tensor<T> TensorSqrbf(Tensor<T> tensor, double beta)
Tensor<T> TensorSqrt(Tensor<T> tensor)
Tensor<T> TensorSqueeze(Tensor<T> tensor, int axis)
Tensor<T> TensorStack(Tensor<T>[] tensors, int axis)
Tensor<T> TensorStd(Tensor<T> tensor)
Tensor<T> TensorSubtract(Tensor<T> a, Tensor<T> b)
Tensor<T> TensorSubtractScalar(Tensor<T> tensor, T scalar)
Tensor<T> TensorSumToSize(Tensor<T> tensor, int[] size)
Tensor<T> TensorSurrogateSpike(Tensor<T> membranePotential, double threshold, double surrogateBeta)
Tensor<T> TensorSwapAxes(Tensor<T> tensor, int axis1, int axis2)
Tensor<T> TensorTake(Tensor<T> tensor, Tensor<int> indices)
Tensor<T> TensorTakeAlongDim(Tensor<T> tensor, Tensor<int> indices, int dim)
Tensor<T> TensorTan(Tensor<T> tensor)
Tensor<T> TensorTanh(Tensor<T> tensor)
Tensor<T> TensorThreshold(Tensor<T> tensor, T threshold, T value)
Tensor<T> TensorThresholdedRelu(Tensor<T> tensor, double threshold)
Tensor<T> TensorTile(Tensor<T> tensor, int[] multiples)
Tensor<T> TensorTopKSoftmax(Tensor<T> scores, int k)
Tensor<T> TensorTranspose(Tensor<T> tensor)
Tensor<T> TensorTrapezoid(Tensor<T> y, double dx, int dim)
Tensor<T> TensorTridiagonalSolve(Tensor<T> lower, Tensor<T> diagonal, Tensor<T> upper, Tensor<T> rhs)
Tensor<T> TensorTril(Tensor<T> tensor, int diagonal)
Tensor<T> TensorTrilIndices(int row, int col, int offset)
Tensor<T> TensorTriu(Tensor<T> tensor, int diagonal)
Tensor<T> TensorTriuIndices(int row, int col, int offset)
Tensor<T> TensorTrunc(Tensor<T> tensor)
Tensor<T> TensorTruncatedNormalFromRng(Random generator, int[] shape, BoxMullerVariant variant, double mean, double stdDev, double low, double high)
Tensor<T> TensorUnflatten(Tensor<T> tensor, int dim, int[] sizes)
Tensor<T> TensorUnfold(Tensor<T> tensor, int dim, int size, int step)
Tensor<T> TensorUniform(int[] shape, double low, double high, int? seed)
Tensor<T> TensorUnique(Tensor<T> input, bool sorted)
Tensor<T> TensorUniqueConsecutive(Tensor<T> input)
Tensor<T> TensorUniqueDim(Tensor<T> tensor, int dim)
Tensor<T> TensorUpsampleBilinear(Tensor<T> input, int[] outputSize)
Tensor<T> TensorVStack(Tensor<T>[] tensors)
Tensor<T> TensorVar(Tensor<T> tensor)
Tensor<T> TensorViewAs(Tensor<T> tensor, Tensor<T> other)
Tensor<T> TensorViewAsReal(Tensor<Complex<T>> input)
Tensor<T> TensorWhere(Tensor<Bit> condition, Tensor<T> x, Tensor<T> y)
Tensor<T> TensorWhere(Tensor<T> condition, Tensor<T> x, Tensor<T> y)
Tensor<T> TensorXlog1py(Tensor<T> x, Tensor<T> y)
Tensor<T> TensorXlogy(Tensor<T> x, Tensor<T> y)
Tensor<T> TensorZerosLike(Tensor<T> tensor)
Tensor<T> TensorZeta(Tensor<T> x, Tensor<T> q)
Tensor<T> ThresholdBackward(Tensor<T> gradOutput, Tensor<T> input, double threshold)
Tensor<T> TileAxisGpu(Tensor<T> input, int axis, int repeats)
Tensor<T> TileBatchGpu(Tensor<T> input, int repeats)
Tensor<T> TimeStretch(Tensor<T> waveform, double rate, int nFft, int hopLength)
Tensor<T> TopKGpu(Tensor<T> input, int k, Tensor<int> indices, bool sorted)
Tensor<T> TopKWeightedRowMix(Tensor<T> table, Tensor<int> indices, Tensor<T> weights)
Tensor<T> TopKWeightedRowMixBackwardTable(Tensor<T> gradOutput, Tensor<int> indices, Tensor<T> weights, int numEntries)
Tensor<T> TopKWeightedRowMixBackwardWeights(Tensor<T> gradOutput, Tensor<T> table, Tensor<int> indices)
Tensor<T> TransposeGpu(Tensor<T> input)
Tensor<T> TryLstmSequenceTrain(Tensor<T> input, Tensor<T> wIh, Tensor<T> wHh, Tensor<T> bias)
Tensor<T> Unfold(Tensor<T> input, int[] kernelSize, int[] stride, int[] padding)
Tensor<T> UploadToContext(T[] data, int[] shape, GpuTensorRole role)
Tensor<T> UploadToContext(Tensor<T> tensor, GpuTensorRole role)
Tensor<T> UploadToGpu(Tensor<T> tensor, GpuTensorRole role)
Tensor<T> UploadToGpu(float[] data, int[] shape, GpuTensorRole role)
Tensor<T> Upsample(Tensor<T> input, int scaleH, int scaleW)
Tensor<T> UpsampleBackwardGpu(Tensor<T> gradOutput, int inputHeight, int inputWidth, int scaleFactor)
Tensor<T> UpsampleGpu(Tensor<T> input, int scaleFactor)
Tensor<T> VarBackward(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> mean, int[] axes)
Tensor<T> WriteOutputToStableCapture(Tensor<T> output)
Tensor<T> XLstmScanForward(Tensor<T> qProj, Tensor<T> kProj, Tensor<T> vProj, Tensor<T> iGate, Tensor<T> fGate, Tensor<T> oGate, int numHeads)
Tensor<T> ZerosGpu(int[] shape)
Tensor<T> ZerosInContext(int[] shape, GpuTensorRole role)
Tensor<T>[] MultiresolutionHashEncodingBackward(Tensor<T> positions, Tensor<T>[] hashTables, int baseResolution, double perLevelScale, int featuresPerLevel, Tensor<T> outputGradient, HashGridAlignment alignment)
Tensor<T>[] TensorChunk(Tensor<T> tensor, int chunks, int dim)
Tensor<T>[] TensorGradient(Tensor<T> tensor, double spacing, int[] dims)
Tensor<T>[] TensorMeshgrid(Tensor<T>[] tensors, string indexing)
Tensor<T>[] TensorSplit(Tensor<T> tensor, int numSplits, int axis)
Tensor<T>[] TensorSplitWithSizes(Tensor<T> tensor, int[] sizes, int dim)
Tensor<T>[] TensorTensorSplit(Tensor<T> tensor, int[] indices, int dim)
Tensor<T>[] TensorUnravelIndex(Tensor<T> indices, int[] shape)
Tensor<T>[] TensorUnstack(Tensor<T> tensor, int axis)
Tensor<TOut> TensorCast(Tensor<TIn> input, RoundingMode mode, CastOverflowBehavior overflow)
Tensor<float> ArgMaxGpu(Tensor<T> input, int axis)
Tensor<float> FftGeneric(Tensor<float> input, bool inverse, FftElementType elementType)
Tensor<float> FusedDDIMStep(Tensor<float> xT, Tensor<float> epsilonTheta, float alphaBarT, float alphaBarTMinus1)
Tensor<float> FusedLoRAForward(Tensor<float> input, Tensor<float> baseOutput, Tensor<float> loraA, Tensor<float> loraB, float scaling)
Tensor<float> FusedSparseLinear(Tensor<float> input, int[] sparseRowOffsets, int[] sparseColIndices, Tensor<float> sparseValues, Tensor<float> bias, FusedActivationType activation)
Tensor<int> BatchedNms(Tensor<T> boxes, Tensor<T> scores, Tensor<int> classIds, double iouThreshold)
Tensor<int> MasksToBoxes(Tensor<T> masks)
Tensor<int> MuLawEncoding(Tensor<T> input, int quantizationChannels)
Tensor<int> Nms(Tensor<T> boxes, Tensor<T> scores, double iouThreshold)
Tensor<int> TensorArgsort(Tensor<T> input, int axis, bool descending)
Tensor<int> TensorBinCount(Tensor<int> input, int? minLength)
Tensor<int> TensorBucketize(Tensor<T> input, Tensor<T> boundaries, bool right)
Tensor<int> TensorHistogram(Tensor<T> input, int bins, T min, T max)
Tensor<int> TensorHistogramDD(Tensor<T> samples, int[] bins, T[] mins, T[] maxs)
Tensor<int> TensorLinearSumAssignment(Tensor<T> cost)
Tensor<int> TensorNonzero(Tensor<T> tensor)
Tensor<int> TensorNonzeroStatic(Tensor<T> tensor, int size, int fillValue)
Tensor<int> TensorSearchSorted(Tensor<T> sortedSequence, Tensor<T> values, bool right)
ValueTuple<T, T> TensorAminmax(Tensor<T> tensor)
ValueTuple<T, int> TensorKthvalue(Tensor<T> input, int k)
ValueTuple<T, int> TensorMode(Tensor<T> input)
ValueTuple<Tensor<T>, Tensor<T>, Tensor<T>> BatchNormBackwardGpu(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> gamma, Tensor<T> saveMean, Tensor<T> saveInvVar, float epsilon)
ValueTuple<Tensor<T>, Tensor<T>, Tensor<T>> FusedBatchNormGpu(Tensor<T> input, Tensor<T> gamma, Tensor<T> beta, Tensor<T> runningMean, Tensor<T> runningVar, double epsilon, double momentum, bool training)
ValueTuple<Tensor<T>, Tensor<T>, Tensor<T>> LayerNormBackwardGpu(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> gamma, Tensor<T> saveMean, Tensor<T> saveInvVar, double epsilon)
ValueTuple<Tensor<T>, Tensor<T>, Tensor<T>> LayerNormGpu(Tensor<T> input, Tensor<T> gamma, Tensor<T> beta, double epsilon)
ValueTuple<Tensor<T>, Tensor<T>, Tensor<T>> ScaledDotProductAttentionBackwardGpu(Tensor<T> gradOutput, Tensor<T> query, Tensor<T> key, Tensor<T> value, Tensor<T> attentionWeights, double scale, bool isCausal)
ValueTuple<Tensor<T>, Tensor<T>, Tensor<bool>, Tensor<T>> SampleRaysWithOccupancy(Tensor<T> rayOrigins, Tensor<T> rayDirections, Tensor<uint> occupancyBitfield, int gridSize, Vector<T> sceneBoundsMin, Vector<T> sceneBoundsMax, T nearBound, T farBound, int maxSamples)
ValueTuple<Tensor<T>, Tensor<T>> BoxIouBackward(Tensor<T> gradOutput, Tensor<T> boxesA, Tensor<T> boxesB)
ValueTuple<Tensor<T>, Tensor<T>> CompleteBoxIouBackward(Tensor<T> gradOutput, Tensor<T> boxesA, Tensor<T> boxesB)
ValueTuple<Tensor<T>, Tensor<T>> DistanceBoxIouBackward(Tensor<T> gradOutput, Tensor<T> boxesA, Tensor<T> boxesB)
ValueTuple<Tensor<T>, Tensor<T>> DropoutGpu(Tensor<T> input, float dropoutRate, bool isTraining, ulong seed)
ValueTuple<Tensor<T>, Tensor<T>> DynamicRoutingGpu(Tensor<T> predictions, int numIterations, float epsilon)
ValueTuple<Tensor<T>, Tensor<T>> GeneralizedBoxIouBackward(Tensor<T> gradOutput, Tensor<T> boxesA, Tensor<T> boxesB)
ValueTuple<Tensor<T>, Tensor<T>> PReLUBackward(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> alpha)
ValueTuple<Tensor<T>, Tensor<T>> TensorCTCForwardBackward(Tensor<T> logProbs, Tensor<int> targets, int[] inputLengths, int[] targetLengths, int blank)
ValueTuple<Tensor<T>, Tensor<T>> TensorLstmCell(Tensor<T> input, Tensor<T> hidden, Tensor<T> cell, Tensor<T> wIh, Tensor<T> wHh, Tensor<T> bIh, Tensor<T> bHh)
ValueTuple<Tensor<T>, Tensor<T>> TensorMeshgrid(Tensor<T> x, Tensor<T> y)
ValueTuple<Tensor<T>, Tensor<T>> TensorRecurrent(RnnCellType cell, Tensor<T> input, Tensor<T> h0, IReadOnlyList<Tensor<T>> weights, bool hasBiases, int numLayers, double dropout, bool training, bool bidirectional, bool batchFirst)
ValueTuple<Tensor<T>, Tensor<T>> TensorStdMean(Tensor<T> tensor, int[] axes, int correction, bool keepDims)
ValueTuple<Tensor<T>, Tensor<T>> TensorVarMean(Tensor<T> tensor, int[] axes, int correction, bool keepDims)
ValueTuple<Tensor<T>, Tensor<int>, Tensor<int>> TensorUniqueConsecutiveWithInfo(Tensor<T> input, bool returnInverse, bool returnCounts)
ValueTuple<Tensor<T>, Tensor<int>, Tensor<int>> TensorUniqueWithInfo(Tensor<T> input, bool sorted, bool returnInverse, bool returnCounts)
ValueTuple<Tensor<T>, Tensor<int>> TensorAdaptiveMaxPoolWithIndices(Tensor<T> input, int[] outputSize)
ValueTuple<Tensor<T>, Tensor<int>> TensorFractionalMaxPool(Tensor<T> input, int[] kernelSize, int[] outputSize, int? seed)
ValueTuple<Tensor<T>, Tensor<int>> TensorFrexp(Tensor<T> tensor)
ValueTuple<Tensor<T>, Tensor<int>> TensorMaxPool1DWithIndices(Tensor<T> input, int kernelSize, int stride, int padding, int dilation, bool ceilMode)
ValueTuple<Tensor<T>, Tensor<int>> TensorSort(Tensor<T> input, int axis, bool descending)
bool InvalidateWeightCache(T[] data)
bool IsDeviceAuthoritative(Tensor<T> tensor)
bool ShouldUseGpu(int elementCount)
bool SupportsFftElementType(FftElementType type)
bool TensorAllClose(Tensor<T> a, Tensor<T> b, T rtol, T atol, bool equalNan)
bool TensorEqual(Tensor<T> a, Tensor<T> b)
bool TryAccumulateSumOfSquares(Tensor<T> tensor, Tensor<T> scalarAccumulator)
bool TryConcatenateChannelsResidentInto(Tensor<T> output, Tensor<T> a, Tensor<T> b)
bool WithGpuContext(Action<GpuExecutionContext> action, GpuExecutionOptions options)
byte[] TensorToBytes(Tensor<T> input, NumericStorageType targetType, bool littleEndian)
float PowerIterationGpu(Tensor<T> weights, Tensor<T> u, Tensor<T> v, int numIterations, float epsilon)
float[] DownloadResidentBuffer(Tensor<T> t)
int ReleaseUnusedDeviceMemory()
int TensorCountNonzero(Tensor<T> tensor)
void ClearActionScratchPool()
void ClearActivationCache()
void CopyGpu(Tensor<T> source, Tensor<T> destination, int size)
void CopyGpu(Tensor<T> source, int srcOffset, Tensor<T> destination, int destOffset, int size)
void DestroyGpuGraph(IntPtr graphExec)
void Dispose()
void DropActivationCache()
void DropAllWeightCaches()
void EnsureResidentInput(Tensor<T> t)
void FillGpu(Tensor<T> buffer, float value, int size)
void FusedDDIMStepInto(Tensor<float> output, Tensor<float> xT, Tensor<float> epsilonTheta, float alphaBarT, float alphaBarTMinus1)
void FusedLoRAForwardInto(Tensor<float> output, Tensor<float> input, Tensor<float> baseOutput, Tensor<float> loraA, Tensor<float> loraB, float scaling)
void FusedSparseLinearInto(Tensor<float> output, Tensor<float> input, int[] sparseRowOffsets, int[] sparseColIndices, Tensor<float> sparseValues, Tensor<float> bias, FusedActivationType activation)
void GridSampleBackwardGpu(Tensor<T> gradOutput, Tensor<T> input, Tensor<T> grid, Tensor<T> gradInput, Tensor<T> gradGrid, int paddingMode, bool alignCorners)
void InvalidateAllWeightCaches()
void InvalidatePersistentTensor(Tensor<T> tensor)
void InvalidateResidentWeightBuffer(Tensor<T> tensor)
void LaunchGpuGraph(IntPtr graphExec)
void LeakyReLUInPlace(Tensor<T> tensor, T alpha)
void NativeComplexPhaseCoherenceDecode(ReadOnlySpan<T> codesRe, ReadOnlySpan<T> codesIm, ReadOnlySpan<T> queryRe, ReadOnlySpan<T> queryIm, Span<T> scores, int V, int D)
void NativeComplexPointwiseMultiply(ReadOnlySpan<T> aRe, ReadOnlySpan<T> aIm, ReadOnlySpan<T> bRe, ReadOnlySpan<T> bIm, Span<T> cRe, Span<T> cIm, bool conjugateB)
void NativeGather(ReadOnlySpan<T> input, ReadOnlySpan<int> indices, Span<T> output)
void NativeHRRBindAccumulate(ReadOnlySpan<T> keyCodeRe, ReadOnlySpan<T> keyCodeIm, ReadOnlySpan<T> valPermCodeRe, ReadOnlySpan<T> valPermCodeIm, ReadOnlySpan<int> keyIds, ReadOnlySpan<int> valIds, Span<T> memoryRe, Span<T> memoryIm, int D)
void NativeUnitPhaseCodebook(Span<T> outRe, Span<T> outIm, int seed, int V, int D, bool kPsk, int k)
void PrepareStableCaptureOutput(Tensor<T> sampleOutput)
void ProjectGaussians3DTo2D(Tensor<T> means3D, Tensor<T> covariances3D, Matrix<T> viewMatrix, Matrix<T> projMatrix, int imageWidth, int imageHeight, Tensor<T> means2D, Tensor<T> covariances2D, Tensor<T> depths, Tensor<bool> visible)
void ReLUInPlace(Tensor<T> tensor)
void RefreshResidentInputInPlace(Tensor<T> t)
void RegisterPersistentTensor(Tensor<T> tensor, PersistentTensorRole role)
void SgdMomentumUpdateGpu(Tensor<T> param, Tensor<T> gradient, Tensor<T> velocity, float learningRate, float momentum, float weightDecay)
void SigmoidInPlace(Tensor<T> tensor)
void SinCos(Vector<T> vector, Vector<T> sinResult, Vector<T> cosResult)
void StdpUpdateGpu(Tensor<T> weights, Tensor<T> preTrace, Tensor<T> postTrace, Tensor<T> preSpike, Tensor<T> postSpike, double ltpRate, double ltdRate, double homeostasisRate, double minWeight, double maxWeight)
void SwishInPlace(Tensor<T> tensor)
void SynchronizeStream()
void TensorAddInPlace(Tensor<T> a, Tensor<T> b)
void TensorLogInto(Tensor<T> destination, Tensor<T> input)
void TensorMultiplyInPlace(Tensor<T> a, Tensor<T> b)
void TensorPermuteInto(Tensor<T> output, Tensor<T> tensor, int[] axes)
void TensorRandomUniformFromGeneratorInto(Random generator, Tensor<T> destination, double min, double max)
void TensorScatterRowsInPlace(Tensor<T> table, Tensor<long> rows, Tensor<T> source)
void TensorSubtractInPlace(Tensor<T> a, Tensor<T> b)
void UnregisterPersistentTensor(Tensor<T> tensor)
void UpdateTracesGpu(Tensor<T> traces, Tensor<T> spikes, Tensor<T> input, float decay, float threshold)