org.nd4j.linalg.api.ops.impl.layers.convolution.Conv3D 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.convolution;
import lombok.Builder;
import lombok.Getter;
import lombok.NonNull;
import lombok.extern.slf4j.Slf4j;
import lombok.val;
import org.nd4j.autodiff.samediff.SDVariable;
import org.nd4j.autodiff.samediff.SameDiff;
import org.nd4j.common.base.Preconditions;
import org.nd4j.imports.NoOpNameFoundException;
import org.nd4j.imports.converters.DifferentialFunctionClassHolder;
import org.nd4j.imports.descriptors.properties.AttributeAdapter;
import org.nd4j.imports.descriptors.properties.PropertyMapping;
import org.nd4j.imports.descriptors.properties.adapters.*;
import org.nd4j.imports.graphmapper.tf.TFGraphMapper;
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.impl.layers.convolution.config.Conv3DConfig;
import org.nd4j.linalg.api.ops.impl.layers.convolution.config.PaddingMode;
import org.nd4j.linalg.util.LinAlgExceptions;
import org.tensorflow.framework.AttrValue;
import org.tensorflow.framework.GraphDef;
import org.tensorflow.framework.NodeDef;
import java.lang.reflect.Field;
import java.util.*;
@Slf4j
@Getter
public class Conv3D extends DynamicCustomOp {
protected Conv3DConfig config;
private static final String INVALID_CONFIGURATION = "Invalid Conv3D configuration : sW = %s pH = %s dW = %s ";
public Conv3D() {
}
public Conv3D(@NonNull SameDiff sameDiff, @NonNull SDVariable input, @NonNull SDVariable weights,
SDVariable bias, @NonNull Conv3DConfig config) {
this(sameDiff, wrapFilterNull(input, weights, bias), config);
}
@Builder(builderMethodName = "sameDiffBuilder")
public Conv3D(SameDiff sameDiff, SDVariable[] inputFunctions, Conv3DConfig config) {
super(sameDiff, inputFunctions);
initConfig(config);
}
public Conv3D(INDArray[] inputs, INDArray[] outputs, Conv3DConfig config){
super(inputs, outputs);
initConfig(config);
}
public Conv3D(@NonNull INDArray input, @NonNull INDArray weights, INDArray bias, INDArray output, @NonNull Conv3DConfig config){
this(wrapFilterNull(input, weights, bias), wrapOrNull(output), config);
}
public Conv3D(INDArray input, INDArray weights, INDArray bias, Conv3DConfig config) {
this(wrapFilterNull(input, weights, bias), null, config);
}
public Conv3D(INDArray input, INDArray weights, Conv3DConfig config) {
this(wrapFilterNull(input, weights), null, config);
}
private void initConfig(Conv3DConfig config){
this.config = config;
Preconditions.checkState(config.getSW() >= 1 && config.getPH() >= 0 && config.getDW() >= 1,
INVALID_CONFIGURATION,
config.getSW(), config.getPH(), config.getDW());
addArgs();
}
@Override
public void configureFromArguments() {
if(config == null && iArguments.size() >= 14) {
Conv3DConfig.Conv3DConfigBuilder builder = Conv3DConfig.builder();
builder.kD(getIArgument(0));
builder.kH(getIArgument(1));
builder.kW(getIArgument(2));
builder.sD(getIArgument(3));
builder.sH(getIArgument(4));
builder.sW(getIArgument(5));
builder.pD(getIArgument(6));
builder.pH(getIArgument(7));
builder.pW(getIArgument(8));
builder.dD(getIArgument(9));
builder.dH(getIArgument(10));
builder.dW(getIArgument(11));
builder.paddingMode(PaddingMode.fromNumber(getIArgument(12).intValue()));
builder.dataFormat(getIArgument(13) > 0 ? "NCDHW" : "NCHDWC");
this.config = builder.build();
}
}
@Override
public void setPropertiesForFunction(Map properties) {
if(config == null) {
Conv3DConfig.Conv3DConfigBuilder builder = Conv3DConfig.builder();
Long dD = getLongValueFromProperty("dD",properties);
if(dD != null)
builder.dD(dD);
Long dH = getLongValueFromProperty("dH",properties);
if(dH != null)
builder.dH(dH);
Long dW = getLongValueFromProperty("dW",properties);
if(dW != null)
builder.dW(dW);
Long sW = getLongValueFromProperty("sW",properties);
if(sW != null)
builder.sW(sW);
Long sD = getLongValueFromProperty("sD",properties);
if(sD != null)
builder.sD(sD);
Long sH = getLongValueFromProperty("sH",properties);
if(sH != null)
builder.sH(sH);
Long pW = getLongValueFromProperty("pW",properties);
if(pW != null)
builder.pW(pW);
Long pD = getLongValueFromProperty("pD",properties);
if(pD != null)
builder.pD(pD);
Long pH = getLongValueFromProperty("pH",properties);
if(pH != null)
builder.pH(pH);
Long kD = getLongValueFromProperty("kD",properties);
if(kD != null)
builder.kD(kD);
Long kW = getLongValueFromProperty("kW",properties);
if(kW != null)
builder.kW(kW);
Long kH = getLongValueFromProperty("kH",properties);
if(kH != null)
builder.kH(kH);
Boolean biasUsed = getBooleanFromProperty("biasUsed",properties);
if(biasUsed != null)
builder.biasUsed(biasUsed);
if(properties.containsKey("dataFormat")) {
builder.dataFormat(properties.get("dataFormat").toString());
}
if(properties.containsKey("paddingMode")) {
builder.paddingMode(PaddingMode.VALID.valueOf(properties.get("paddingMode").toString()));
}
this.config = builder.build();
}
}
private void addArgs() {
if(getConfig().getPaddingMode() == null)
getConfig().setPaddingMode(PaddingMode.VALID);
addIArgument(
// TODO: support bias terms
// ArrayUtil.fromBoolean(getConfig().isBiasUsed()),
getConfig().getKD(),
getConfig().getKH(),
getConfig().getKW(),
getConfig().getSD(),
getConfig().getSH(),
getConfig().getSW(),
getConfig().getPD(),
getConfig().getPH(),
getConfig().getPW(),
getConfig().getDD(),
getConfig().getDH(),
getConfig().getDW(),
getConfig().getPaddingMode().index,
getConfig().isNCDHW() ? 0 : 1
);
}
@Override
public Object getValue(Field property) {
if (config == null && !iArguments.isEmpty()) {
LinAlgExceptions.assertAllConfigured(this,12);
createConfigFromArgs();
}
return config.getValue(property);
}
public void createConfigFromArgs() {
config = Conv3DConfig.builder()
.kD(iArguments.get(0))
.kH(iArguments.get(1))
.kW(iArguments.get(2))
.sD(iArguments.get(3))
.sH(iArguments.get(4))
.sW(iArguments.get(5))
.pD(iArguments.get(6))
.pH(iArguments.get(7))
.pW(iArguments.get(8))
.dD(iArguments.get(9))
.dH(iArguments.get(10))
.dW(iArguments.get(11))
.paddingMode(PaddingMode.fromNumber(iArguments.get(12).intValue()))
.dataFormat(iArguments.get(13) == 1 ? Conv3DConfig.NCDHW : Conv3DConfig.NDHWC)
.build();
}
@Override
public long[] iArgs() {
if (iArguments.size() == 0)
addArgs();
return super.iArgs();
}
@Override
public Map> attributeAdaptersForFunction() {
Map> ret = new LinkedHashMap<>();
Map tfAdapters = new LinkedHashMap<>();
val fields = DifferentialFunctionClassHolder.getInstance().getFieldsForFunction(this);
//TF uses [kD, kH, kW, iC, oC] for weights
tfAdapters.put("kD", new NDArrayShapeAdapter(0));
tfAdapters.put("kH", new NDArrayShapeAdapter(1));
tfAdapters.put("kW", new NDArrayShapeAdapter(2));
tfAdapters.put("sD", new IntArrayIntIndexAdapter(1));
tfAdapters.put("sH", new IntArrayIntIndexAdapter(2));
tfAdapters.put("sW", new IntArrayIntIndexAdapter(3));
tfAdapters.put("pD", new IntArrayIntIndexAdapter(1));
tfAdapters.put("pH", new IntArrayIntIndexAdapter(2));
tfAdapters.put("pW", new IntArrayIntIndexAdapter(3));
tfAdapters.put("isSameMode", new StringNotEqualsAdapter("VALID"));
ret.put(tensorflowName(), tfAdapters);
return ret;
}
@Override
public Map propertiesForFunction() {
if (config == null) {
return Collections.emptyMap();
}
return config.toProperties();
}
@Override
public String opName() {
return "conv3dnew";
}
@Override
public Map> mappingsForFunction() {
Map> ret = new HashMap<>();
Map map = new HashMap<>();
val kernelMapping = PropertyMapping.builder()
.propertyNames(new String[]{"kD", "kW", "kH"})
.tfInputPosition(1)
.onnxAttrName("kernel_shape")
.build();
val strideMapping = PropertyMapping.builder()
.tfAttrName("strides")
.onnxAttrName("strides")
.propertyNames(new String[]{"sD", "sW", "sH"})
.build();
val dilationMapping = PropertyMapping.builder()
.onnxAttrName("dilations")
.propertyNames(new String[]{"dD", "dH", "dW"})
.tfAttrName("rates")
.build();
val sameMode = PropertyMapping.builder()
.onnxAttrName("auto_pad")
.propertyNames(new String[]{"isSameMode"})
.tfAttrName("padding")
.build();
val paddingWidthHeight = PropertyMapping.builder()
.onnxAttrName("padding")
.propertyNames(new String[]{"pD", "pW", "pH"})
.build();
val dataFormat = PropertyMapping.builder()
.onnxAttrName("data_format")
.tfAttrName("data_format")
.propertyNames(new String[]{"dataFormat"})
.build();
val outputPadding = PropertyMapping.builder()
.propertyNames(new String[]{"aD", "aH", "aW"})
.build();
val biasUsed = PropertyMapping.builder()
.propertyNames(new String[]{"biasUsed"})
.build();
for (val propertyMapping : new PropertyMapping[]{
kernelMapping,
strideMapping,
dilationMapping,
sameMode,
paddingWidthHeight,
dataFormat,
outputPadding, biasUsed}) {
for (val keys : propertyMapping.getPropertyNames())
map.put(keys, propertyMapping);
}
ret.put(tensorflowName(), map);
return ret;
}
@Override
public void initFromTensorFlow(NodeDef nodeDef, SameDiff initWith, Map attributesForNode, GraphDef graph) {
TFGraphMapper.initFunctionFromProperties(nodeDef.getOp(), this, attributesForNode, nodeDef, graph);
addArgs();
}
@Override
public List doDiff(List f1) {
List ret = new ArrayList<>();
List inputs = new ArrayList<>();
inputs.addAll(Arrays.asList(args()));
inputs.add(f1.get(0));
if (config == null && !iArguments.isEmpty()) {
LinAlgExceptions.assertAllConfigured(this,12);
createConfigFromArgs();
}
Conv3DDerivative conv3DDerivative = Conv3DDerivative.derivativeBuilder()
.conv3DConfig(config)
.inputFunctions(inputs.toArray(new SDVariable[inputs.size()]))
.sameDiff(sameDiff)
.build();
ret.addAll(Arrays.asList(conv3DDerivative.outputVariables()));
return ret;
}
@Override
public boolean isConfigProperties() {
return true;
}
@Override
public String configFieldName() {
return "config";
}
@Override
public String onnxName() {
throw new NoOpNameFoundException("No ONNX op name found for: " + getClass().getName());
}
@Override
public String tensorflowName() {
return "Conv3D";
}
@Override
public List calculateOutputDataTypes(List inputDataTypes){
int n = args().length;
Preconditions.checkState(inputDataTypes != null && inputDataTypes.size() == n, "Expected %s input data types for %s, got %s", n, getClass(), inputDataTypes);
return Collections.singletonList(inputDataTypes.get(0));
}
}