org.nd4j.linalg.api.ops.impl.transforms.strict.PreciseGELU Maven / Gradle / Ivy
/*******************************************************************************
* Copyright (c) 2015-2018 Skymind, Inc.
*
* This program and the accompanying materials are made available under the
* terms of the Apache License, Version 2.0 which is available at
* https://www.apache.org/licenses/LICENSE-2.0.
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* License for the specific language governing permissions and limitations
* under the License.
*
* SPDX-License-Identifier: Apache-2.0
******************************************************************************/
package org.nd4j.linalg.api.ops.impl.transforms.strict;
import org.nd4j.autodiff.samediff.SDVariable;
import org.nd4j.autodiff.samediff.SameDiff;
import org.nd4j.imports.NoOpNameFoundException;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.api.ops.BaseTransformStrictOp;
import java.util.Collections;
import java.util.List;
/**
* GELU activation function - Gaussian Error Linear Units
* For more details, see Gaussian Error Linear Units (GELUs) - https://arxiv.org/abs/1606.08415
* Note: This op implements both the sigmoid and tanh-based approximations; to use the sigmoid approximation (recommended)
* use precise=false; otherwise, use precise = true for the slower but marginally more accurate tanh version.
* @author [email protected]
*/
public class PreciseGELU extends BaseTransformStrictOp {
public PreciseGELU(SameDiff sameDiff, SDVariable i_v, boolean inPlace, boolean precise) {
super(sameDiff, i_v, inPlace);
}
public PreciseGELU() {
}
public PreciseGELU(INDArray x, INDArray z) {
super(x, z);
}
public PreciseGELU(INDArray ndArray) {
super(ndArray);
}
@Override
public int opNum() {
return 55;
}
@Override
public String opName() {
return "precise_gelu";
}
@Override
public String onnxName() {
throw new NoOpNameFoundException("No onnx op opName found for " + opName());
}
@Override
public String tensorflowName() {
throw new NoOpNameFoundException("Tensorflow name not found for " + opName());
//return "PreciseGELU";
}
@Override
public List doDiff(List i_v) {
SDVariable ret = f().geluDerivative(arg(), true).mul(i_v.get(0));
return Collections.singletonList(ret);
}
}