public static class FusedBatchNormGrad.Inputs<T extends TNumber,U extends TNumber> extends RawOpInputs<FusedBatchNormGrad<T,U>>
Modifier and Type | Field and Description |
---|---|
String |
dataFormat
The data format for y_backprop, x, x_backprop.
|
float |
epsilon
A small float number added to the variance of x.
|
boolean |
isTraining
A bool value to indicate the operation is for training (default)
or inference.
|
Operand<U> |
reserveSpace1
When is_training is True, a 1D Tensor for the computed batch
mean to be reused in gradient computation.
|
Operand<U> |
reserveSpace2
When is_training is True, a 1D Tensor for the computed batch
variance (inverted variance in the cuDNN case) to be reused in
gradient computation.
|
Operand<U> |
reserveSpace3
When is_training is True, a 1D Tensor for some intermediate results to be reused
in gradient computation.
|
Operand<TFloat32> |
scale
A 1D Tensor for scaling factor, to scale the normalized x.
|
DataType |
T
The data type for the elements of input and output Tensors.
|
DataType |
U
The data type for the scale, offset, mean, and variance.
|
Operand<T> |
x
A 4D Tensor for input data.
|
Operand<T> |
yBackprop
A 4D Tensor for the gradient with respect to y.
|
Constructor and Description |
---|
Inputs(GraphOperation op) |
attributeMetadata, attributeNames, attributes, attributeValue, attributeValues, equals, getOutputs, hashCode, toString
public final Operand<T extends TNumber> yBackprop
public final Operand<TFloat32> scale
public final Operand<U extends TNumber> reserveSpace1
public final Operand<U extends TNumber> reserveSpace2
public final Operand<U extends TNumber> reserveSpace3
public final DataType T
public final DataType U
public final float epsilon
public final String dataFormat
public final boolean isTraining
public Inputs(GraphOperation op)
Copyright © 2015–2022. All rights reserved.