$OpDocMathAbs
$OpDocMathAbs
Input tensor.
Result as a new tensor.
$OpDocMathAcos
$OpDocMathAcos
Input tensor.
Result as a new tensor.
$OpDocMathAcosh
$OpDocMathAcosh
Input tensor.
Result as a new tensor.
$OpDocMathAdd
$OpDocMathAdd
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathAddN
$OpDocMathAddN
Input tensors.
Result as a new tensor.
$OpDocMathAll
$OpDocMathAll
Input tensor to reduce.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
If true
, retain the reduced axes.
Result as a new tensor.
$OpDocMathAny
$OpDocMathAny
Input tensor to reduce.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
If true
, retain the reduced axes.
Result as a new tensor.
$OpDocMathApproximatelyEqual
$OpDocMathApproximatelyEqual
First input tensor.
Second input tensor.
Comparison tolerance value.
Result as a new tensor.
$OpDocMathArgmax
$OpDocMathArgmax
Input tensor.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
Data type for the output tensor.
Result as a new tensor.
$OpDocMathArgmax
$OpDocMathArgmax
Input tensor.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
Result as a new tensor.
$OpDocMathArgmin
$OpDocMathArgmin
Input tensor.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
Data type for the output tensor.
Result as a new tensor.
$OpDocMathArgmin
$OpDocMathArgmin
Input tensor.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
Result as a new tensor.
$OpDocMathAsin
$OpDocMathAsin
Input tensor.
Result as a new tensor.
$OpDocMathAsinh
$OpDocMathAsinh
Input tensor.
Result as a new tensor.
$OpDocMathAtan
$OpDocMathAtan
Input tensor.
Result as a new tensor.
$OpDocMathAtan2
$OpDocMathAtan2
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathAtanh
$OpDocMathAtanh
Input tensor.
Result as a new tensor.
$OpDocMathBinCount
$OpDocMathBinCount
Tensor containing non-negative values.
If not null
, this tensor must have the same shape as input
. For each value in input
, the
corresponding bin count will be incremented by the corresponding weight instead of 1
.
If not null
, this ensures the output has length at least minLength
, padding with zeros at
the end, if necessary.
If not null
, this skips values in input
that are equal or greater than maxLength
,
ensuring that the output has length at most maxLength
.
If weights
is null
, this determines the data type used for the output tensor (i.e., the
tensor containing the bin counts).
Result as a new tensor.
$OpDocMathBucketize
$OpDocMathBucketize
Numeric tensor to bucketize.
Sorted sequence of numbers specifying the boundaries of the buckets.
Result as a new tensor.
$OpDocMathCeil
$OpDocMathCeil
Input tensor.
Result as a new tensor.
$OpDocMathComplex
$OpDocMathComplex
Tensor containing the real component.
Tensor containing the imaginary component.
Result as a new tensor.
$OpDocMathComplex
$OpDocMathComplex
Tensor containing the real component.
Tensor containing the imaginary component.
Result as a new tensor.
$OpDocMathConjugate
$OpDocMathConjugate
Input tensor.
Result as a new tensor.
$OpDocMathCos
$OpDocMathCos
Input tensor.
Result as a new tensor.
$OpDocMathCosh
$OpDocMathCosh
Input tensor.
Result as a new tensor.
$OpDocMathCountNonZero
$OpDocMathCountNonZero
Input tensor to reduce.
Integer array containing the axes to reduce. If null
, then all axes are reduced.
If true
, retain the reduced axes.
Result as a new tensor.
$OpDocMathCross
$OpDocMathCross
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathCumprod
$OpDocMathCumprod
Input tensor.
INT32
tensor containing the axis along which to perform the cumulative product.
Boolean value indicating whether to perform an exclusive cumulative product.
Boolean value indicating whether to perform a reverse cumulative product.
Result as a new tensor.
$OpDocMathCumsum
$OpDocMathCumsum
Input tensor.
Tensor containing the axis along which to perform the cumulative sum.
Boolean value indicating whether to perform an exclusive cumulative sum.
Boolean value indicating whether to perform a reverse cumulative sum.
Result as a new tensor.
$OpDocMathDiag
$OpDocMathDiag
Diagonal values, represented as a rank-K
tensor, where K
can be at most 3
.
Result as a new tensor.
$OpDocMathDiagPart
$OpDocMathDiagPart
Rank-K
input tensor, where K
is either 2
, 4
, or 6
.
Result as a new tensor.
$OpDocMathDigamma
$OpDocMathDigamma
Input tensor.
Result as a new tensor.
$OpDocMathDivide
$OpDocMathDivide
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathEqual
$OpDocMathEqual
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathErf
$OpDocMathErf
Input tensor.
Result as a new tensor.
$OpDocMathErfc
$OpDocMathErfc
Input tensor.
Result as a new tensor.
$OpDocMathExp
$OpDocMathExp
Input tensor.
Result as a new tensor.
$OpDocMathExpm1
$OpDocMathExpm1
Input tensor.
Result as a new tensor.
$OpDocMathFloor
$OpDocMathFloor
Input tensor.
Result as a new tensor.
$OpDocMathFloorMod
$OpDocMathFloorMod
First input tensor.
Second input tensor.
Result as a new tensor.
OpDocMathGreater
OpDocMathGreater
First input tensor.
Second input tensor.
Result as a new tensor.
OpDocMathGreaterEqual
OpDocMathGreaterEqual
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathIgamma
$OpDocMathIgamma
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathIgammac
$OpDocMathIgammac
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathIncompleteBeta
$OpDocMathIncompleteBeta
First input tensor.
Second input tensor.
Third input tensor.
Result as a new tensor.
$OpDocMathIsFinite
$OpDocMathIsFinite
Input tensor.
Result as a new tensor.
$OpDocMathIsInf
$OpDocMathIsInf
Input tensor.
Result as a new tensor.
$OpDocMathIsNaN
$OpDocMathIsNaN
Input tensor.
Result as a new tensor.
OpDocMathLess
OpDocMathLess
First input tensor.
Second input tensor.
Result as a new tensor.
OpDocMathLessEqual
OpDocMathLessEqual
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathLinspace
$OpDocMathLinspace
Rank 0 (i.e., scalar) tensor that contains the starting value of the number sequence.
Rank 0 (i.e., scalar) tensor that contains the ending value (inclusive) of the number sequence.
Rank 0 (i.e., scalar) tensor that contains the number of values in the number sequence.
Result as a new tensor.
$OpDocMathLog
$OpDocMathLog
Input tensor.
Result as a new tensor.
$OpDocMathLog1p
$OpDocMathLog1p
Input tensor.
Result as a new tensor.
$OpDocMathLogGamma
$OpDocMathLogGamma
Input tensor.
Result as a new tensor.
$OpDocMathLogSigmoid
$OpDocMathLogSigmoid
Input tensor.
Result as a new tensor.
$OpDocMathLogSumExp
$OpDocMathLogSumExp
Input tensor to reduce.
Integer sequence containing the axes to reduce. If null
, then all axes are reduced.
If true
, retain the reduced axes.
Result as a new tensor.
$OpDocMathLogicalAnd
$OpDocMathLogicalAnd
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathLogicalNot
$OpDocMathLogicalNot
Input tensor.
Result as a new tensor.
$OpDocMathLogicalOr
$OpDocMathLogicalOr
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathLogicalXOr
$OpDocMathLogicalXOr
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathMatmul
$OpDocMathMatmul
First input tensor.
Second input tensor.
If true
, a
is transposed before the multiplication.
If true
, b
is transposed before the multiplication.
If true
, a
is conjugated before the multiplication.
If true
, b
is conjugated before the multiplication.
If true
, a
is treated as a sparse matrix (i.e., it is assumed it contains many zeros).
If true
, b
is treated as a sparse matrix (i.e., it is assumed it contains many zeros).
Result as a new tensor.
$OpDocMathMatrixBandPart
$OpDocMathMatrixBandPart
Input tensor.
Scalar tensor that contains the number of sub-diagonals to keep. If negative, the entire lower triangle is kept.
Scalar tensor that contains the number of super-diagonals to keep. If negative, the entire upper triangle is kept.
Result as a new tensor containing the expected banded tensor and has rank K
and same shape as input
.
$OpDocMathMatrixDiag
$OpDocMathMatrixDiag
Rank-K
input tensor, where K >= 1
.
Result as a new tensor with rank equal to K + 1
and shape equal to the shape of diagonal
, with its
last dimension duplicated.
$OpDocMathMatrixDiagPart
$OpDocMathMatrixDiagPart
Rank-K
tensor, where K >= 2
.
Result as a new tensor containing the diagonal(s) and having shape equal to
input.shape[:-2] + [min(input.shape[-2:])]
.
$OpDocMathMatrixSetDiag
$OpDocMathMatrixSetDiag
Rank-K+1
tensor, where K >= 2
.
Rank-K
tensor, where K >= 1
.
Result as a new tensor with rank equal to K + 1
and shape equal to the shape of input
.
$OpDocMathMax
$OpDocMathMax
Input tensor to reduce.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
If true
, retain the reduced axes.
Result as a new tensor.
$OpDocMathMaximum
$OpDocMathMaximum
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathMean
$OpDocMathMean
Input tensor to reduce.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
If true
, retain the reduced axes.
Result as a new tensor.
$OpDocMathMin
$OpDocMathMin
Input tensor to reduce.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
If true
, retain the reduced axes.
Result as a new tensor.
$OpDocMathMinimum
$OpDocMathMinimum
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathMod
$OpDocMathMod
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathMultiply
$OpDocMathMultiply
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathNegate
$OpDocMathNegate
Input tensor.
Result as a new tensor.
$OpDocMathNotEqual
$OpDocMathNotEqual
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathPolygamma
$OpDocMathPolygamma
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathPow
$OpDocMathPow
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathProd
$OpDocMathProd
Input tensor to reduce.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
If true
, retain the reduced axes.
Result as a new tensor.
$OpDocMathRange
$OpDocMathRange
Rank 0 (i.e., scalar) tensor that contains the starting value of the number sequence.
Rank 0 (i.e., scalar) tensor that contains the ending value (exclusive) of the number sequence.
Rank 0 (i.e., scalar) tensor that contains the difference between consecutive numbers in the sequence.
Result as a new tensor.
$OpDocMathRealDivide
$OpDocMathRealDivide
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathReciprocal
$OpDocMathReciprocal
Input tensor.
Result as a new tensor.
$OpDocMathRound
$OpDocMathRound
Input tensor.
Result as a new tensor.
$OpDocMathRoundInt
$OpDocMathRoundInt
Input tensor.
Result as a new tensor.
$OpDocMathRsqrt
$OpDocMathRsqrt
Input tensor.
Result as a new tensor.
$OpDocMathScalarMul
$OpDocMathScalarMul
Scalar tensor.
Tensor to multiply the scalar tensor with.
Result as a new tensor.
$OpDocMathSegmentMax
$OpDocMathSegmentMax
Data (must have a numeric data type -- i.e., representing a number).
Segment indices. Values should be sorted and can be repeated.
Result as a new tensor.
$OpDocMathSegmentMean
$OpDocMathSegmentMean
Data (must have a numeric data type -- i.e., representing a number).
Segment indices. Values should be sorted and can be repeated.
Result as a new tensor.
$OpDocMathSegmentMin
$OpDocMathSegmentMin
Data (must have a numeric data type -- i.e., representing a number).
Segment indices. Values should be sorted and can be repeated.
Result as a new tensor.
$OpDocMathSegmentProd
$OpDocMathSegmentProd
Data (must have a numeric data type -- i.e., representing a number).
Segment indices. Values should be sorted and can be repeated.
Result as a new tensor.
$OpDocMathSegmentSum
$OpDocMathSegmentSum
Data (must have a numeric data type -- i.e., representing a number).
Segment indices. Values should be sorted and can be repeated.
Result as a new tensor.
$OpDocMathSelect
$OpDocMathSelect
Boolean condition tensor.
Tensor which may have the same shape as condition
. If condition
has rank 1
, then t
may
have a higher rank, but its first dimension must match the size of condition
.
Tensor with the same data type and shape as t
.
Result as a new tensor.
$OpDocMathSigmoid
$OpDocMathSigmoid
Input tensor.
Result as a new tensor.
$OpDocMathSign
$OpDocMathSign
Input tensor.
Result as a new tensor.
$OpDocMathSin
$OpDocMathSin
Input tensor.
Result as a new tensor.
$OpDocMathSinh
$OpDocMathSinh
Input tensor.
Result as a new tensor.
$OpDocMathSparseSegmentMean
$OpDocMathSparseSegmentMean
Data (must have a numeric data type -- i.e., representing a number).
One-dimensional tensor with rank equal to that of segmentIndices
.
Segment indices. Values should be sorted and can be repeated.
Optional scalar indicating the size of the output tensor.
Result as a new tensor.
$OpDocMathSparseSegmentSum
$OpDocMathSparseSegmentSum
Data (must have a numeric data type -- i.e., representing a number).
One-dimensional tensor with rank equal to that of segmentIndices
.
Segment indices. Values should be sorted and can be repeated.
Optional scalar indicating the size of the output tensor.
Result as a new tensor.
$OpDocMathSparseSegmentSumSqrtN
$OpDocMathSparseSegmentSumSqrtN
Data (must have a numeric data type -- i.e., representing a number).
One-dimensional tensor with rank equal to that of segmentIndices
.
Segment indices. Values should be sorted and can be repeated.
Optional scalar indicating the size of the output tensor.
Result as a new tensor.
$OpDocMathSqrt
$OpDocMathSqrt
Input tensor.
Result as a new tensor.
$OpDocMathSquare
$OpDocMathSquare
Input tensor.
Result as a new tensor.
$OpDocMathSquaredDifference
$OpDocMathSquaredDifference
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathSubtract
$OpDocMathSubtract
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathSum
$OpDocMathSum
Input tensor to reduce.
Integer tensor containing the axes to reduce. If null
, then all axes are reduced.
If true
, retain the reduced axes.
Result as a new tensor.
$OpDocMathTan
$OpDocMathTan
Input tensor.
Result as a new tensor.
$OpDocMathTanh
$OpDocMathTanh
Input tensor.
Result as a new tensor.
Dynamic version (i.e., where axesA
and axesB
may be tensors) of the tensorDot
op.
Dynamic version (i.e., where axesA
and axesB
may be tensors) of the tensorDot
op.
$OpDocMathTensorDot
First tensor.
Second tensor.
Axes to contract in a
.
Axes to contract in b
.
Created op output.
InvalidShapeException
If axesA
or axesB
is not a scalar.
Dynamic version (i.e., where numAxes
may be a tensor) of the tensorDot
op.
Dynamic version (i.e., where numAxes
may be a tensor) of the tensorDot
op.
$OpDocMathTensorDot
First tensor.
Second tensor.
Number of axes to contract.
Created op output.
InvalidShapeException
If numAxes
is not a scalar.
$OpDocMathTrace
$OpDocMathTrace
Input tensor.
Result as a new tensor.
$OpDocMathTruncateDivide
$OpDocMathTruncateDivide
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathTruncateMod
$OpDocMathTruncateMod
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathUnsortedSegmentMax
$OpDocMathUnsortedSegmentMax
Data (must have a numeric data type -- i.e., representing a number).
Segment indices.
Number of segments.
Result as a new tensor.
$OpDocMathUnsortedSegmentSum
$OpDocMathUnsortedSegmentSum
Data (must have a numeric data type -- i.e., representing a number).
Segment indices.
Number of segments.
Result as a new tensor.
$OpDocMathZerosFraction
$OpDocMathZerosFraction
Input tensor.
Result as a new tensor.
$OpDocMathZeta
$OpDocMathZeta
First input tensor.
Second input tensor.
Result as a new tensor.
$OpDocMathFloorDivide
$OpDocMathFloorDivide
First input tensor.
Second input tensor.
Result as a new tensor.
(Since version 0.1) Use truncateDivide
instead.