deeplearning4j/deeplearning4j-nn/src/main/java/org/deeplearning4j/nn/layers/samediff/SameDiffOutputLayer.java
/*
* ******************************************************************************
* *
* *
* * 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.deeplearning4j.nn.layers.samediff;
import lombok.Getter;
import lombok.Setter;
import lombok.val;
import org.deeplearning4j.nn.api.Layer;
import org.deeplearning4j.nn.api.layers.IOutputLayer;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.gradient.DefaultGradient;
import org.deeplearning4j.nn.gradient.Gradient;
import org.deeplearning4j.nn.layers.AbstractLayer;
import org.deeplearning4j.nn.workspace.ArrayType;
import org.deeplearning4j.nn.workspace.LayerWorkspaceMgr;
import org.nd4j.autodiff.samediff.SDVariable;
import org.nd4j.autodiff.samediff.SameDiff;
import org.nd4j.autodiff.samediff.array.SingleThreadArrayHolder;
import org.nd4j.autodiff.samediff.internal.InferenceSession;
import org.nd4j.autodiff.samediff.internal.SessionMemMgr;
import org.nd4j.common.base.Preconditions;
import org.nd4j.linalg.api.buffer.DataType;
import org.nd4j.linalg.api.memory.MemoryWorkspace;
import org.nd4j.linalg.api.memory.conf.WorkspaceConfiguration;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.dataset.api.DataSet;
import org.nd4j.linalg.dataset.api.iterator.DataSetIterator;
import org.nd4j.linalg.factory.Nd4j;
import org.nd4j.common.primitives.Pair;
import java.util.*;
public class SameDiffOutputLayer extends AbstractLayer<org.deeplearning4j.nn.conf.layers.samediff.SameDiffOutputLayer>
implements IOutputLayer {
public static final String INPUT_KEY = "input";
public static final String LABELS_KEY = "labels";
protected SameDiff sameDiff;
protected SDVariable outputVar;
protected String outputKey;
@Getter @Setter
protected INDArray labels;
protected INDArray params;
protected INDArray gradients;
protected Map<String,INDArray> paramTable;
protected Map<String,INDArray> gradTable;
public SameDiffOutputLayer(NeuralNetConfiguration conf, DataType dataType){
super(conf, dataType);
}
@Override
public Layer clone() {
throw new UnsupportedOperationException();
}
@Override
public boolean isPretrainLayer() {
return false;
}
@Override
public void clearNoiseWeightParams() {
//TODO - properly support weight noise...
}
@Override
public INDArray activate(boolean training, LayerWorkspaceMgr workspaceMgr) {
return activateHelper(true, workspaceMgr);
}
private INDArray activateHelper(boolean activations, LayerWorkspaceMgr workspaceMgr){
assertInputSet(false);
//Check where the output occurs. If it's a simple loss layer (no params) this could
// just be the input!
if(activations && INPUT_KEY.equals(layerConf().activationsVertexName())){
return workspaceMgr.leverageTo(ArrayType.ACTIVATIONS, input);
}
//TODO optimize
try(MemoryWorkspace ws = Nd4j.getWorkspaceManager().scopeOutOfWorkspaces()) {
if (sameDiff == null) {
doInit();
}
}
//Configure memory management for SameDiff instance - use DL4J workspaces
String wsNameWorking = workspaceMgr.getWorkspaceName(ArrayType.FF_WORKING_MEM);
String wsNameOutput = workspaceMgr.getWorkspaceName(ArrayType.ACTIVATIONS);
WorkspaceConfiguration confWorking = workspaceMgr.getConfiguration(ArrayType.FF_WORKING_MEM);
WorkspaceConfiguration confOutput = workspaceMgr.getConfiguration(ArrayType.ACTIVATIONS);
boolean actScopedOut = workspaceMgr.isScopedOut(ArrayType.ACTIVATIONS);
Preconditions.checkState(actScopedOut || wsNameOutput != null, "Activations must have a workspace or must be scoped out");
SessionMemMgr mmgr = new DL4JSameDiffMemoryMgr(wsNameWorking, wsNameOutput, confWorking, confOutput);
InferenceSession is = sameDiff.getSessions().get(Thread.currentThread().getId());
if(is == null){
is = SameDiff.getInferenceFactory().create(sameDiff);
sameDiff.getSessions().put(Thread.currentThread().getId(), is);
}
is.setMmgr(mmgr);
Map<String,INDArray> phMap = new HashMap<>();
phMap.put(INPUT_KEY, input);
if(!activations && layerConf().labelsRequired() && labels != null) {
phMap.put(LABELS_KEY, labels);
}
String s = activations ? layerConf().activationsVertexName() : outputVar.name();
INDArray out = sameDiff.outputSingle(phMap, s);
//Clear placeholders and op inputs to ensure no out-of-scope arrays are still referenced anywhere
sameDiff.clearPlaceholders(true);
sameDiff.clearOpInputs();
//Edge case: vertex is just an Identity function, for example
//TODO there may be a cleaner way to do this...
if(!actScopedOut && !out.data().getParentWorkspace().getId().equals(wsNameOutput)){
out = workspaceMgr.dup(ArrayType.ACTIVATIONS, out);
} else if(actScopedOut && out.isAttached()){
out = out.detach();
}
return out;
}
@Override
public Pair<Gradient, INDArray> backpropGradient(INDArray epsilon, LayerWorkspaceMgr workspaceMgr) {
assertInputSet(true);
Preconditions.checkState(!layerConf().labelsRequired() || labels != null, "Cannot execute backprop: Labels are not set. " +
"If labels are not required for this SameDiff output layer, override SameDiffOutputLayer.labelsRequired()" +
" to return false instead");
Gradient g = new DefaultGradient();
INDArray dLdIn;
try(MemoryWorkspace ws = Nd4j.getWorkspaceManager().scopeOutOfWorkspaces()) {
if (sameDiff == null) {
//Usually doInit will be called in forward pass; not necessarily the case in output layers
// (for efficiency, we skip output layer forward pass in MultiLayerNetwork/ComputationGraph)
doInit();
}
if(sameDiff.getFunction("grad") == null)
sameDiff.createGradFunction(INPUT_KEY);
}
//Configure memory management for SameDiff instance - use DL4J workspaces
Map<Long,InferenceSession> sessionMap = sameDiff.getFunction("grad").getSessions();
if(!sessionMap.containsKey(Thread.currentThread().getId())){
sessionMap.put(Thread.currentThread().getId(), SameDiff.getInferenceFactory().create(sameDiff.getFunction("grad")));
}
String wsNameWorking = workspaceMgr.getWorkspaceName(ArrayType.BP_WORKING_MEM);
String wsNameActGrad = workspaceMgr.getWorkspaceName(ArrayType.ACTIVATION_GRAD);
WorkspaceConfiguration confWorking = workspaceMgr.getConfiguration(ArrayType.BP_WORKING_MEM);
WorkspaceConfiguration confOutput = workspaceMgr.getConfiguration(ArrayType.ACTIVATION_GRAD);
boolean actGradScopedOut = workspaceMgr.isScopedOut(ArrayType.ACTIVATION_GRAD);
Preconditions.checkState(actGradScopedOut || wsNameActGrad != null, "Activation gradients must have a workspace or be scoped out");
SessionMemMgr mmgr = new DL4JSameDiffMemoryMgr(wsNameWorking, wsNameActGrad, confWorking, confOutput);
sessionMap.get(Thread.currentThread().getId()).setMmgr(mmgr);
if(!sameDiff.hasGradientFunction()) {
//Create when scoped out, to ensure any arrays are not in WS
sameDiff.createGradFunction(INPUT_KEY);
}
List<String> gradVarNames = new ArrayList<>();
gradVarNames.addAll(paramTable.keySet());
gradVarNames.add(INPUT_KEY);
Map<String,INDArray> phMap = new HashMap<>();
phMap.put(INPUT_KEY, input);
phMap.put(LABELS_KEY, labels);
Map<String,INDArray> grads = sameDiff.calculateGradients(phMap, gradVarNames);
for(String s : paramTable.keySet() ){
INDArray sdGrad = grads.get(s);
INDArray dl4jGrad = gradTable.get(s);
dl4jGrad.assign(sdGrad); //TODO OPTIMIZE THIS
g.gradientForVariable().put(s, dl4jGrad);
if(sdGrad.closeable()){
sdGrad.close();
}
}
dLdIn = grads.get(INPUT_KEY);
//Clear placeholders and op inputs to ensure no out-of-scope arrays are still referenced anywhere
sameDiff.clearPlaceholders(true);
sameDiff.clearOpInputs();
//TODO there may be a cleaner way to do this...
if(!actGradScopedOut && !dLdIn.data().getParentWorkspace().getId().equals(wsNameActGrad)){
dLdIn = workspaceMgr.dup(ArrayType.ACTIVATION_GRAD, dLdIn);
} else if(actGradScopedOut && dLdIn.isAttached()){
dLdIn = dLdIn.detach();
}
return new Pair<>(g, dLdIn);
}
/**Returns the parameters of the neural network as a flattened row vector
* @return the parameters of the neural network
*/
@Override
public INDArray params() {
return params;
}
@Override
public INDArray getParam(String param) {
return paramTable.get(param);
}
@Override
public long numParams(){
return params == null ? 0 : (int)params.length();
}
@Override
public void setParam(String key, INDArray val) {
if(!paramTable.containsKey(key)){
throw new IllegalArgumentException("Cannot set parameter, invalid/unknown parameter key: " + key);
}
INDArray current = paramTable.get(key);
if(!Arrays.equals(current.shape(), val.shape())){
throw new IllegalArgumentException("Cannot set parameter \"" + key + "\", invalid shape: parameter array has shape "
+ Arrays.toString(current.shape()) + ", trying to set parameter of shape " + Arrays.toString(val.shape()));
}
}
@Override
public void setParams(INDArray params) {
if (params != null) {
throw new UnsupportedOperationException("Not supported");
}
}
protected void setParams(INDArray params, char order) {
setParams(params);
}
@Override
public void setParamsViewArray(INDArray params) {
this.params = params;
}
@Override
public INDArray getGradientsViewArray() {
return gradients;
}
@Override
public void setBackpropGradientsViewArray(INDArray gradients) {
this.gradients = gradients;
this.gradTable = layerConf().initializer().getGradientsFromFlattened(conf(), gradients);
}
@Override
public void setParamTable(Map<String, INDArray> paramTable) {
if(this.paramTable == null){
this.paramTable = paramTable;
} else {
for (Map.Entry<String, INDArray> e : paramTable.entrySet()) {
setParam(e.getKey(), e.getValue());
}
}
}
@Override
public Map<String, INDArray> paramTable() {
return paramTable(false);
}
@Override
public Map<String, INDArray> paramTable(boolean backpropParamsOnly) {
return paramTable;
}
protected void doInit(){
try(MemoryWorkspace ws = Nd4j.getWorkspaceManager().scopeOutOfWorkspaces()) {
org.deeplearning4j.nn.conf.layers.samediff.SameDiffOutputLayer bl = layerConf();
sameDiff = SameDiff.create();
//Use SingleThreadArrayHolder so we can use views (also don't nede multithreading here, DL4J is not thread safe)
sameDiff.setArrayHolders(new SingleThreadArrayHolder(), new SingleThreadArrayHolder(), false);
Map<String, INDArray> p = paramTable();
long[] inputShape = input.shape().clone();
inputShape[0] = -1;
SDVariable inputVar = sameDiff.placeHolder(INPUT_KEY, dataType, inputShape);
SDVariable labelVar = null;
if(layerConf().labelsRequired()){
long[] labelShape = labels == null ? new long[]{-1, -1} : labels.shape().clone();
labelShape[0] = -1;
labelVar = sameDiff.placeHolder(LABELS_KEY, dataType, labelShape);
}
Map<String, long[]> paramShapes = layerConf().getLayerParams().getParamShapes();
Map<String, SDVariable> params = new LinkedHashMap<>();
for (String s : paramShapes.keySet()) {
val ps = paramShapes.get(s);
SDVariable v = sameDiff.var(s, dataType, ps);
params.put(s, v);
}
SDVariable layerOutput = bl.defineLayer(sameDiff, inputVar, labelVar, params);
Preconditions.checkNotNull(layerOutput, "Invalid output: layer output is null");
outputVar = layerOutput;
for (Map.Entry<String, INDArray> e : p.entrySet()) {
INDArray arr = e.getValue();
sameDiff.associateArrayWithVariable(arr, sameDiff.getVariable(e.getKey()));
}
this.outputKey = layerOutput.name();
}
}
@Override
public boolean needsLabels() {
return layerConf().labelsRequired();
}
@Override
public double computeScore(double fullNetRegTerm, boolean training, LayerWorkspaceMgr workspaceMgr) {
INDArray scoreArr = activateHelper(false, workspaceMgr);
return (scoreArr.getDouble(0) + fullNetRegTerm) / input.size(0);
}
@Override
public INDArray computeScoreForExamples(double fullNetRegTerm, LayerWorkspaceMgr workspaceMgr) {
throw new UnsupportedOperationException("Not yet implemented");
}
@Override
public double f1Score(DataSet data) {
throw new UnsupportedOperationException("Not supported");
}
@Override
public double f1Score(INDArray examples, INDArray labels) {
throw new UnsupportedOperationException("Not supported");
}
@Override
public int numLabels() {
return 0;
}
@Override
public void fit(DataSetIterator iter) {
throw new UnsupportedOperationException("Not supported");
}
@Override
public int[] predict(INDArray examples) {
throw new UnsupportedOperationException("Not supported");
}
@Override
public List<String> predict(DataSet dataSet) {
throw new UnsupportedOperationException("Not supported");
}
@Override
public void fit(INDArray examples, INDArray labels) {
throw new UnsupportedOperationException("Not supported");
}
@Override
public void fit(DataSet data) {
throw new UnsupportedOperationException("Not supported");
}
@Override
public void fit(INDArray examples, int[] labels) {
throw new UnsupportedOperationException("Not supported");
}
}