Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance. Project price only 1 $
You can buy this project and download/modify it how often you want.
/*******************************************************************************
* 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.deeplearning4j.nn.conf.preprocessor;
import lombok.*;
import org.deeplearning4j.nn.api.MaskState;
import org.deeplearning4j.nn.conf.InputPreProcessor;
import org.deeplearning4j.nn.conf.inputs.InputType;
import org.deeplearning4j.util.TimeSeriesUtils;
import org.nd4j.base.Preconditions;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.api.shape.Shape;
import org.nd4j.linalg.primitives.Pair;
import org.nd4j.linalg.util.ArrayUtil;
import org.deeplearning4j.nn.workspace.LayerWorkspaceMgr;
import org.deeplearning4j.nn.workspace.ArrayType;
import org.nd4j.shade.jackson.annotation.JsonCreator;
import org.nd4j.shade.jackson.annotation.JsonProperty;
import java.util.Arrays;
/**A preprocessor to allow CNN and RNN layers to be used together.
* For example, ConvolutionLayer -> GravesLSTM
* Functionally equivalent to combining CnnToFeedForwardPreProcessor + FeedForwardToRnnPreProcessor
* Specifically, this does two things:
* (a) Reshape 4d activations out of CNN layer, with shape [timeSeriesLength*miniBatchSize, numChannels, inputHeight, inputWidth])
* into 3d (time series) activations (with shape [numExamples, inputHeight*inputWidth*numChannels, timeSeriesLength])
* for use in RNN layers
* (b) Reshapes 3d epsilons (weights.*deltas) out of RNN layer (with shape
* [miniBatchSize,inputHeight*inputWidth*numChannels,timeSeriesLength]) into 4d epsilons with shape
* [miniBatchSize*timeSeriesLength, numChannels, inputHeight, inputWidth] suitable to feed into CNN layers.
* Note: numChannels is equivalent to channels or featureMaps referenced in different literature
* @author Alex Black
*/
@Data
@EqualsAndHashCode(exclude = {"product"})
public class CnnToRnnPreProcessor implements InputPreProcessor {
private long inputHeight;
private long inputWidth;
private long numChannels;
@Getter(AccessLevel.NONE)
@Setter(AccessLevel.NONE)
private long product;
@JsonCreator
public CnnToRnnPreProcessor(@JsonProperty("inputHeight") long inputHeight,
@JsonProperty("inputWidth") long inputWidth, @JsonProperty("numChannels") long numChannels) {
this.inputHeight = inputHeight;
this.inputWidth = inputWidth;
this.numChannels = numChannels;
this.product = inputHeight * inputWidth * numChannels;
}
@Override
public INDArray preProcess(INDArray input, int miniBatchSize, LayerWorkspaceMgr workspaceMgr) {
if (input.rank() != 4)
throw new IllegalArgumentException(
"Invalid input: expect CNN activations with rank 4 (received input with shape "
+ Arrays.toString(input.shape()) + ")");
if(input.size(1) != numChannels || input.size(2) != inputHeight || input.size(3) != inputWidth){
throw new IllegalStateException("Invalid input, does not match configuration: expected [minibatch, numChannels="
+ numChannels + ", inputHeight=" + inputHeight + ", inputWidth=" + inputWidth + "] but got input array of" +
"shape " + Arrays.toString(input.shape()));
}
//Input: 4d activations (CNN)
//Output: 3d activations (RNN)
if (input.ordering() != 'c' || !Shape.hasDefaultStridesForShape(input))
input = input.dup('c');
val shape = input.shape(); //[timeSeriesLength*miniBatchSize, numChannels, inputHeight, inputWidth]
//First: reshape 4d to 2d, as per CnnToFeedForwardPreProcessor
INDArray twod = input.reshape('c', input.size(0), ArrayUtil.prod(input.shape()) / input.size(0));
//Second: reshape 2d to 3d, as per FeedForwardToRnnPreProcessor
INDArray reshaped = workspaceMgr.dup(ArrayType.ACTIVATIONS, twod, 'f');
reshaped = reshaped.reshape('f', miniBatchSize, shape[0] / miniBatchSize, product);
return reshaped.permute(0, 2, 1);
}
@Override
public INDArray backprop(INDArray output, int miniBatchSize, LayerWorkspaceMgr workspaceMgr) {
if (output.ordering() == 'c' || !Shape.hasDefaultStridesForShape(output))
output = output.dup('f');
val shape = output.shape();
INDArray output2d;
if (shape[0] == 1) {
//Edge case: miniBatchSize = 1
output2d = output.tensorAlongDimension(0, 1, 2).permutei(1, 0);
} else if (shape[2] == 1) {
//Edge case: timeSeriesLength = 1
output2d = output.tensorAlongDimension(0, 1, 0);
} else {
//As per FeedForwardToRnnPreprocessor
INDArray permuted3d = output.permute(0, 2, 1);
output2d = permuted3d.reshape('f', shape[0] * shape[2], shape[1]);
}
if (shape[1] != product)
throw new IllegalArgumentException("Invalid input: expected output size(1)=" + shape[1]
+ " must be equal to " + inputHeight + " x columns " + inputWidth + " x channels "
+ numChannels + " = " + product + ", received: " + shape[1]);
INDArray ret = workspaceMgr.dup(ArrayType.ACTIVATION_GRAD, output2d, 'c');
return ret.reshape('c', output2d.size(0), numChannels, inputHeight, inputWidth);
}
@Override
public CnnToRnnPreProcessor clone() {
return new CnnToRnnPreProcessor(inputHeight, inputWidth, numChannels);
}
@Override
public InputType getOutputType(InputType inputType) {
if (inputType == null || inputType.getType() != InputType.Type.CNN) {
throw new IllegalStateException("Invalid input type: Expected input of type CNN, got " + inputType);
}
InputType.InputTypeConvolutional c = (InputType.InputTypeConvolutional) inputType;
val outSize = c.getChannels() * c.getHeight() * c.getWidth();
return InputType.recurrent(outSize);
}
@Override
public Pair feedForwardMaskArray(INDArray maskArray, MaskState currentMaskState,
int minibatchSize) {
//Assume mask array is 4d - a mask array that has been reshaped from [minibatch,timeSeriesLength] to [minibatch*timeSeriesLength, 1, 1, 1]
if (maskArray == null) {
return new Pair<>(maskArray, currentMaskState);
} else {
//Need to reshape mask array from [minibatch*timeSeriesLength, 1, 1, 1] to [minibatch,timeSeriesLength]
return new Pair<>(TimeSeriesUtils.reshapeCnnMaskToTimeSeriesMask(maskArray, minibatchSize),currentMaskState);
}
}
}
