Abs returns the Op that represents the output of the corresponding operation.

Add returns the element-wise sum of the two values.

And returns the element-wise logic "and" operator.

ArgMinMax calculates the "argmin" or "argmax" across an axis of the given input array x.

BatchNormForInference implements Batch Norm for inference.

BatchNormForTraining implements Batch Norm for training.

BatchNormGradient calculates the BatchNorm gradient.

Broadcast prefixes dimensions to an array by duplicating the data in the array.

BroadcastInDim broadcasts x to an output with the given shape.

Call will evaluate a subComputation with the given operands.

Ceil returns the Op that represents the output of the corresponding operation.

Clz returns element-wise the "count leading zeros" bits of input node x -- for integer values.

Complex returns the complex number taking x0 as the real part and x1 as the imaginary part.

Concatenate results on the given axis.

Conj returns the conjugate of a complex number.

Constant introduces an Op.

ConvertDType of x to dtype.

ConvGeneralDilated is a generic Convolution operation offered by XLA.

Cos returns the Op that represents the output of the corresponding operation.

DecodeArgMinMax retrieves the arguments for a ArgMinMax op.

DecodeBatchNormForInference retrieves the arguments for the BatchNormForInference op.

DecodeBatchNormForTraining retrieves the arguments for the BatchNormForTraining op.

DecodeBatchNormGrad retrieves the arguments for the BatchNormGradient op.

DecodeBroadcast retrieves the arguments for a Broadcast op.

DecodeBroadcastInDim retrieves the arguments for a BroadcastInDim op.

DecodeConcatenate retrieves the arguments for a Concatenate op.

DecodeConvertDType retrieves the arguments for a ConvertDType op.

DecodeConvGeneralDilated retrieves the arguments for the ConvGeneralDilated op.

DecodeDotGeneral retrieves the arguments for a DotGeneral op.

DecodeDynamicSlice retrieves the arguments for the DynamicSlice op.

DecodeDynamicUpdateSlice retrieves the arguments for the DynamicUpdateSlice op.

DecodeFFT retrieves the arguments for the FFT op.

DecodeGather retrieves the arguments for a Gather op.

DecodeGetTupleElement retrieves the arguments of an GetTupleElement op.

DecodeIota retrieves the arguments of an Iota op.

DecodePad retrieves the arguments for a Pad op.

DecodeParameter extracts the arguments to the Parameter call that created the op.

DecodeReduceWindow retrieves the arguments for a ReduceWindow op.

DecodeReshape retrieves the arguments for a Reshape op.

DecodeReverse retrieves the arguments for the Reverse op.

DecodeRngBitGenerator retrieves the arguments for the FFT op.

DecodeScatter retrieves the arguments for a Scatter (ScatterCustom or ScatterAdd) op.

DecodeSelectAndScatter retrieves the arguments for a SelectAndScatter (ScatterAndScatterCustom or ScatterAndScatterMax) op.

DecodeSlice retrieves the arguments for a Slice op.

DecodeTranspose retrieves the arguments for a Transpose op.

DecodeWhile retrieves the arguments for the While op.

Div returns the element-wise subtraction of the two values.

Dot returns the "dot product" operation.

DotGeneral takes as input lhs (left-hand-side) and rhs (right-hand-side) specifications for a general vector product -- a generalized "Einsum".

DynamicSlice extracts a sub-array from the input array at dynamic start_indices.

DynamicUpdateSlice generates a result which is the value of the input array operand, with a slice update overwritten at startIndices.

Equal performs element-wise equality check, returns boolean results with the same dimensions as input.

EqualTotalOrder returns the element-wise operation.

Erf returns the "error function", defined as erf(x) = 2/Pi * \int_{0}^{x}{e^{-t^2}dt}.

Exp returns the Op that represents the output of the corresponding operation.

Expm1 returns the Op that represents the output of the corresponding operation.

FFT calls the XLA FFT operation, which implements {Forward, Inverse} x {Complex, Real} versions.

Floor returns the Op that represents the output of the corresponding operation.

Gather is a powerful but cumbersome Gather operation offered by XLA.

GetTupleElement extracts one element from a Tuple.

GreaterOrEqual performs element-wise comparison, returns boolean results with the same dimensions as input.

GreaterOrEqualTotalOrder returns the element-wise operation.

GreaterThan performs element-wise comparison, returns boolean results with the same dimensions as input.

GreaterThanTotalOrder returns the element-wise operation.

Identity returns an Op whose output is the same as its input.

Imag returns the imaginary part of a complex number.

Iota creates a constant of the given shape with increasing numbers (starting from 0) on the given axis.

IsFinite tests whether each element of operand is finite, i.e., is not positive or negative infinity, and is not NaN.

LessOrEqual performs element-wise comparison, returns boolean results with the same dimensions as input.

LessOrEqualTotalOrder returns the element-wise operation.

LessThan performs element-wise comparison, returns boolean results with the same dimensions as input.

LessThanTotalOrder returns the element-wise operation.

Log returns the Op that represents the output of the corresponding operation.

Log1p returns the expression log(x+1).

LogicalNot returns the Op that represents the output of the corresponding operation.

Logistic returns the element-wise expression 1/(1+exp(-x)).

Max returns the element-wise highest value among the two.

Min returns the element-wise smallest value among the two.

Mul returns the element-wise multiplication of the two values.

Neg returns the Op that represents the output of the corresponding operation.

NotEqual performs element-wise inequality check, returns boolean results with the same dimensions as input.

NotEqualTotalOrder returns the element-wise operation.

Or returns the element-wise logic "and" operator.

Pad injects padding on the start, end or interior (in between each element) of the given operand.

Parameter creates a "retrieves a parameter value" op in builder.

PopulationCount computes the number of bits set in each element of operand.

Pow returns the Op that represents the output of the corresponding operation.

Real return the real part of a complex number.

Reduce the selected axes of the input array, using the given custom reduceComputation.

ReduceAnd is a shortcut for Reduce with the proper computation and initial value to reduce x on the given axes, by taking the logical-and of the reduced axes.

ReduceMax is a shortcut for Reduce with the proper computation and initial value to reduce x on the given axes, by taking the max value.

ReduceMin is a shortcut for Reduce with the proper computation and initial value to reduce x on the given axes, by taking the min value.

ReduceOr is a shortcut for Reduce with the proper computation and initial value to reduce x on the given axes, by taking the logical-or of the reduced axes.

ReduceProduct is a shortcut for Reduce with the proper computation and initial value to reduce x on the given axes, by taking the product of the reduced axes.

ReduceSum is a shortcut for Reduce with the proper computation and initial value to reduce x on the given axes, by taking the sum of the reduced axes.

ReduceWindow applies a reduction function to all elements in each window of x, producing an N multidimensional array as output.

Rem returns the remainder operation, also known as modulo (or Mod for short).

Reshape reshapes x to the new dimensions.

Reverse returns x with the values for the given dimensions reversed, that is, the value indexed at `i` will be swapped with the value at indexed `(dimension_size - 1 - i)`.

RngBitGenerator generates the given shape filled with random bits.

Round returns the Op that represents the output of the corresponding operation.

Rsqrt returns the element-wise reciprocal of square root operation 1/sqrt(x).

ScalarOne returns a one (1) constant for the given dtype.

ScalarZero returns a zero constant for the given dtype.

ScatterAdd values from updates pointed by scatterIndices to operand.

ScatterCustom is a powerful but cumbersome Scatter operation offered by XLA.

ScatterMax scatter values from updates pointed by scatterIndices to operand, by taking the Max.

ScatterMin scatter values from updates pointed by scatterIndices to operand, by taking the Min.

SelectAndScatterCustom runs windows (similar to ReduceWindow) over the operand, selects values (selectComputation) to updates the output (like Scatter) using the scatterComputation with values from source.

SelectAndScatterMax calls SelectAndScatterCustom with a zero defaultValue, sum for updateComputation and an appropriate selectComputation to implement a SelectAndScatter that updates the max value in the windows.

SelectAndScatterMin calls SelectAndScatterCustom with a zero defaultValue, sum for updateComputation and an appropriate selectComputation to implement a SelectAndScatter that updates the max value in the windows.

SelectAndScatterSum calls SelectAndScatterCustom with a zero defaultValue, sum for updateComputation and a selectComputation that always selects.

Sign returns element-wise +1, +/-0 or -1 depending on the sign of x.

Sin returns the Op that represents the output of the corresponding operation.

Slice extracts a sub-array from the input array.

SplitTuple is a convenience wrapper around GetTupleElement, it will return an array with all the nodes.

Sqrt returns the Op that represents the output of the corresponding operation.

Sub returns the element-wise subtraction of the two values.

Tanh returns the Op that represents the output of the corresponding operation.

Transpose axes of x.

Tuple organizes multiple nodes in one tuple-node.

Where takes element-wise values from onTrue or onFalse depending on the value of condition (expected to be boolean).

While executes a loop in the computation.

Xor returns the element-wise logic "and" operator.

ReduceAndType reduces by taking the logical-and value.

ReduceMaxType reduces by taking the maximum value.

ReduceMinType reduces by taking the minimum value.

ReduceOrType reduces by taking the logical-or value.

ReduceProductType reduces by multiplying all elements being reduced.

ReduceSumType reduces by summing all elements being reduced.

UndefinedReduceType is an undefined value.

ConvolveAxesConfig defines the interpretation of the input/kernel/output tensor axes.

PadAxis defines the amount of padding preceding one axis (Start), at the end of axis (End) or in between the inputs (Interior).

OpType enumerates the various operation types supported by XLA.

ReduceOpType select among the basic types of reduction supported, see XlaBuilder.ReduceComputation.