org.nd4j.linalg.api.ops.impl.layers.ExternalErrorsFunction Maven / Gradle / Ivy
The newest version!
/*
* ******************************************************************************
* *
* *
* * 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.
* *
* * See the NOTICE file distributed with this work for additional
* * information regarding copyright ownership.
* * 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.layers;
import onnx.Onnx;
import org.nd4j.autodiff.samediff.SDVariable;
import org.nd4j.autodiff.samediff.SameDiff;
import org.nd4j.autodiff.samediff.VariableType;
import org.nd4j.common.base.Preconditions;
import org.nd4j.imports.NoOpNameFoundException;
import org.nd4j.linalg.api.buffer.DataType;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.api.ops.DynamicCustomOp;
import org.nd4j.linalg.api.ops.Op;
import org.nd4j.linalg.api.ops.OpContext;
import org.nd4j.linalg.api.shape.LongShapeDescriptor;
import org.nd4j.linalg.factory.Nd4j;
import org.tensorflow.framework.AttrValue;
import org.tensorflow.framework.GraphDef;
import org.tensorflow.framework.NodeDef;
import java.util.*;
public class ExternalErrorsFunction extends DynamicCustomOp {
public static final String OP_NAME = "ExternalErrorsFn";
private static final List OUT_SHAPE = Collections.singletonList(LongShapeDescriptor.fromShape(new long[0], Nd4j.dataType()));
private Map gradients;
private Map gradVariables;
private SDVariable out;
private String id;
private String outName;
private List gradVarNames;
public ExternalErrorsFunction(SameDiff sd, List inputs, Map gradients) {
super(sd, inputs.toArray(new SDVariable[inputs.size()]));
if(gradients == null)
gradients = new HashMap<>();
this.gradients = gradients;
gradVarNames = new ArrayList<>();
for(SDVariable input : inputs) {
gradVarNames.add(input.name());
}
this.id = UUID.randomUUID().toString();
}
public ExternalErrorsFunction(){ }
public String getGradPlaceholderName(){
return arg().name() + "-grad";
}
@Override
public void configureWithSameDiff(SameDiff sameDiff) {
super.configureWithSameDiff(sameDiff);
if(outName != null) {
this.out = sameDiff.getVariable(outName);
}
gradients = new HashMap<>();
}
@Override
public Map propertiesForFunction() {
Map ret = new HashMap<>();
if(out != null)
ret.put("out",out);
if(id != null)
ret.put("id",id);
if(gradVarNames != null)
ret.put("gradVarNames",gradVarNames);
return ret;
}
@Override
public void configureFromArguments() {
super.configureFromArguments();
}
@Override
public void setPropertiesForFunction(Map properties) {
if(properties.containsKey("id")) {
this.id = properties.get("id").toString();
}
if(properties.containsKey("out")) {
this.outName = properties.get("out").toString();
}
if(properties.containsKey("gradVarNames")) {
List gradVarNames = (List) properties.get("gradVarNames");
this.gradVarNames = gradVarNames;
}
}
@Override
public SDVariable[] outputVariables(String baseName) {
if(out == null){
if(id == null)
this.id = UUID.randomUUID().toString();
String name = "dummyOutput-" + id;
if(sameDiff.hasVariable(name)){
out = sameDiff.getVariable(name);
} else {
out = sameDiff.zero(name, Nd4j.dataType(), 1);
sameDiff.getOps().get(getOwnName()).setOutputsOfOp(Collections.singletonList(out.name()));
sameDiff.getVariables().get(name).setOutputOfOp(getOwnName());
}
}
return new SDVariable[]{out};
}
@Override
public List doDiff(List f1) {
List out = new ArrayList<>();
if (gradVariables == null) {
gradVariables = new HashMap<>();
for(SDVariable arg : args()){
INDArray gradArr = gradients.get(arg.name());
SDVariable grad;
DataType dt = arg.dataType();
String n = getGradPlaceholderName();
if(gradArr != null){
long[] shape = gradArr.shape().clone();
shape[0] = -1;
grad = sameDiff.var(n, VariableType.PLACEHOLDER, null, dt, shape);
} else {
grad = sameDiff.var(n, VariableType.PLACEHOLDER, null, dt);
}
sameDiff.setGradientForVariableName(arg.name(),grad);
gradVariables.put(arg.name(), grad);
out.add(grad);
}
}
return out;
}
public void updateBeforeExecution() {
Preconditions.checkState(gradVariables != null, "Variables list is null - doDiff has not been called?");
//Update external gradients ready for execution
for(Map.Entry e : gradVariables.entrySet()){
INDArray extGradArray = gradients.get(e.getKey());
if(extGradArray == null){
throw new IllegalStateException("Cannot execute SameDiff instance with external errors: external gradient " +
"for variable " + e.getKey() + " has not been defined");
}
gradVariables.get(e.getKey()).setArray(extGradArray);
}
}
@Override
public void initFromTensorFlow(NodeDef nodeDef, SameDiff initWith, Map attributesForNode, GraphDef graph) {
}
@Override
public void initFromOnnx(Onnx.NodeProto node, SameDiff initWith, Map attributesForNode, Onnx.GraphProto graph) {
}
@Override
public String onnxName() {
throw new NoOpNameFoundException("Not supported: " + opName());
}
@Override
public String tensorflowName() {
throw new NoOpNameFoundException("Not supported: " + opName());
}
@Override
public String opName(){
return OP_NAME;
}
@Override
public String toString(){
return "ExternalErrorsFunction(" + (gradVariables != null ? gradVariables.keySet() : "") + ")";
}
@Override
public List calculateOutputShape(){
return OUT_SHAPE;
}
@Override
public List calculateOutputShape(OpContext oc){
return OUT_SHAPE;
}
public Op.Type opType() {
return Op.Type.LOGIC;
}
}