/**
* Copyright [2012-2014] PayPal Software Foundation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use super file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://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.
*/
package ml.shifu.shifu.core.dtrain.nn;
import java.io.IOException;
import java.util.concurrent.TimeUnit;
import ml.shifu.guagua.ComputableMonitor;
import ml.shifu.guagua.hadoop.io.GuaguaLineRecordReader;
import ml.shifu.guagua.hadoop.io.GuaguaWritableAdapter;
import ml.shifu.guagua.io.GuaguaFileSplit;
import ml.shifu.guagua.util.NumberFormatUtils;
import ml.shifu.guagua.worker.WorkerContext;
import ml.shifu.shifu.container.obj.ColumnConfig;
import ml.shifu.shifu.core.dtrain.dataset.BasicFloatMLData;
import ml.shifu.shifu.core.dtrain.dataset.BasicFloatMLDataPair;
import ml.shifu.shifu.core.dtrain.dataset.FloatMLDataPair;
import org.apache.commons.lang.StringUtils;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
/**
* {@link NNWorker} is used to compute NN model according to splits assigned. The result will be sent to master for
* accumulation.
*
* <p>
* Gradients in each worker will be sent to master to update weights of model in worker, which follows Encog's
* multi-core implementation.
*
* <p>
* {@link NNWorker} is to load data with text format.
*/
@ComputableMonitor(timeUnit = TimeUnit.SECONDS, duration = 300)
public class NNWorker extends AbstractNNWorker<Text> {
@Override
public void load(GuaguaWritableAdapter<LongWritable> currentKey, GuaguaWritableAdapter<Text> currentValue,
WorkerContext<NNParams, NNParams> workerContext) {
super.count += 1;
if((super.count) % 5000 == 0) {
LOG.info("Read {} records.", super.count);
}
float[] inputs = new float[super.subFeatures.size()];
float[] ideal = new float[super.outputNodeCount];
if(super.isDry) {
// dry train, use empty data.
addDataPairToDataSet(0, new BasicFloatMLDataPair(new BasicFloatMLData(inputs), new BasicFloatMLData(ideal)));
return;
}
long hashcode = 0;
float significance = 1f;
// use guava Splitter to iterate only once
// use NNConstants.NN_DEFAULT_COLUMN_SEPARATOR to replace getModelConfig().getDataSetDelimiter(), super follows
// the function in akka mode.
int index = 0, inputsIndex = 0, outputIndex = 0;
for(String input: DEFAULT_SPLITTER.split(currentValue.getWritable().toString())) {
// check here to avoid bad performance in failed NumberFormatUtils.getFloat(input, 0f)
float floatValue = input.length() == 0 ? 0f : NumberFormatUtils.getFloat(input, 0f);
// no idea about why NaN in input data, we should process it as missing value TODO , according to norm type
floatValue = (Float.isNaN(floatValue) || Double.isNaN(floatValue)) ? 0f : floatValue;
if(index == super.columnConfigList.size()) {
// do we need to check if not weighted directly set to 1f; if such logic non-weight at first, then
// weight, how to process???
if(StringUtils.isBlank(modelConfig.getWeightColumnName())) {
significance = 1f;
// break here if we reach weight column which is last column
break;
}
// check here to avoid bad performance in failed NumberFormatUtils.getFloat(input, 1f)
significance = input.length() == 0 ? 1f : NumberFormatUtils.getFloat(input, 1f);
// if invalid weight, set it to 1f and warning in log
if(Float.compare(significance, 0f) < 0) {
LOG.warn("The {} record in current worker weight {} is less than 0f, it is invalid, set it to 1.",
count, significance);
significance = 1f;
}
// the last field is significance, break here
break;
} else {
ColumnConfig columnConfig = super.columnConfigList.get(index);
if(columnConfig != null && columnConfig.isTarget()) {
if(modelConfig.isRegression()) {
ideal[outputIndex++] = floatValue;
} else {
if(modelConfig.getTrain().isOneVsAll()) {
// if one vs all, set correlated idea value according to trainerId which means in trainer
// with id 0, target 0 is treated with 1, other are 0. Such target value are set to index of
// tags like [0, 1, 2, 3] compared with ["a", "b", "c", "d"]
ideal[outputIndex++] = Float.compare(floatValue, trainerId) == 0 ? 1f : 0f;
} else {
int ideaIndex = (int) floatValue;
ideal[ideaIndex] = 1f;
}
}
} else {
if(subFeatureSet.contains(index)) {
inputs[inputsIndex++] = floatValue;
hashcode = hashcode * 31 + Double.valueOf(floatValue).hashCode();
}
}
}
index += 1;
}
// sample negative only logic here
if(modelConfig.getTrain().getSampleNegOnly()) {
if(this.modelConfig.isFixInitialInput()) {
// if fixInitialInput, sample hashcode in 1-sampleRate range out if negative records
int startHashCode = (100 / this.modelConfig.getBaggingNum()) * this.trainerId;
// here BaggingSampleRate means how many data will be used in training and validation, if it is 0.8, we
// should take 1-0.8 to check endHashCode
int endHashCode = startHashCode
+ Double.valueOf((1d - this.modelConfig.getBaggingSampleRate()) * 100).intValue();
if((modelConfig.isRegression() || (modelConfig.isClassification() && modelConfig.getTrain()
.isOneVsAll())) // regression or onevsall
&& (int) (ideal[0] + 0.01d) == 0 // negative record
&& isInRange(hashcode, startHashCode, endHashCode)) {
return;
}
} else {
// if not fixed initial input, and for regression or onevsall multiple classification (regression also).
// if negative record
if((modelConfig.isRegression() || (modelConfig.isClassification() && modelConfig.getTrain()
.isOneVsAll())) // regression or onevsall
&& (int) (ideal[0] + 0.01d) == 0 // negative record
&& Double.compare(super.sampelNegOnlyRandom.nextDouble(),
this.modelConfig.getBaggingSampleRate()) >= 0) {
return;
}
}
}
FloatMLDataPair pair = new BasicFloatMLDataPair(new BasicFloatMLData(inputs), new BasicFloatMLData(ideal));
// up sampling logic, just add more weights while bagging sampling rate is still not changed
if(modelConfig.isRegression() && isUpSampleEnabled() && Double.compare(ideal[0], 1d) == 0) {
// Double.compare(ideal[0], 1d) == 0 means positive tags; sample + 1 to avoid sample count to 0
pair.setSignificance(significance * (super.upSampleRng.sample() + 1));
} else {
pair.setSignificance(significance);
}
boolean isValidation = false;
if(workerContext.getAttachment() != null && workerContext.getAttachment() instanceof Boolean) {
isValidation = (Boolean) workerContext.getAttachment();
}
boolean isInTraining = addDataPairToDataSet(hashcode, pair, isValidation);
// do bagging sampling only for training data，
if(isInTraining) {
float subsampleWeights = sampleWeights(pair.getIdealArray()[0]);
if(isPositive(pair.getIdealArray()[0])) {
this.positiveSelectedTrainCount += subsampleWeights * 1L;
} else {
this.negativeSelectedTrainCount += subsampleWeights * 1L;
}
// set weights to significance, if 0, significance will be 0, that is bagging sampling
pair.setSignificance(pair.getSignificance() * subsampleWeights);
} else {
// for validation data, according bagging sampling logic, we may need to sampling validation data set, while
// validation data set are only used to compute validation error, not to do real sampling is ok.
}
}
/*
* (non-Javadoc)
*
* @see ml.shifu.guagua.worker.AbstractWorkerComputable#initRecordReader(ml.shifu.guagua.io.GuaguaFileSplit)
*/
@Override
public void initRecordReader(GuaguaFileSplit fileSplit) throws IOException {
super.setRecordReader(new GuaguaLineRecordReader(fileSplit));
}
}