org.nd4j.linalg.api.ops.BaseGradientOp Maven / Gradle / Ivy
package org.nd4j.linalg.api.ops;
import org.nd4j.autodiff.samediff.SDVariable;
import org.nd4j.autodiff.samediff.SameDiff;
import org.nd4j.base.Preconditions;
import org.nd4j.linalg.api.ndarray.INDArray;
/**
* A gradient op always makes the following assumptions:
* there is always a y (beacuse of backpropagating
* or using the chain rule)
*
* and that it is special exec (for now)
*
* This op opType sis meant to be used
* to build derivative operations.
*
*
* @author Adam Gibson
*/
public abstract class BaseGradientOp extends BaseTransformOp implements GradientOp {
public BaseGradientOp(SameDiff sameDiff, SDVariable i_v1, SDVariable i_v2) {
super(sameDiff, i_v1, i_v2);
}
public BaseGradientOp(SameDiff sameDiff, SDVariable i_v1, SDVariable i_v2, boolean inPlace) {
super(sameDiff, i_v1, i_v2, inPlace);
}
public BaseGradientOp(INDArray x, INDArray z) {
super(x, z);
assertWrt();
}
public BaseGradientOp() {
}
public BaseGradientOp(INDArray x, INDArray z, long n) {
super(x, z, n);
assertWrt();
}
public BaseGradientOp(INDArray x, INDArray y, INDArray z, long n) {
super(x, y, z, n);
assertWrt();
}
public BaseGradientOp(INDArray x) {
super(x);
assertWrt();
}
/**
* The array
* to the gradient with respect to
*
* @return
*/
@Override
public INDArray wrt() {
return y();
}
@Override
public boolean isExecSpecial() {
return true;
}
@Override
public boolean isPassThrough() {
return true;
}
private void assertWrt() {
Preconditions.checkState(y != null,"A gradient op must define a wrt variable as a Y. ");
}
}