/**
* Takes in two arguments:
*
* <ol>
* <li>The input {@link org.apache.hadoop.fs.Path} where the input documents live
* <li>The output {@link org.apache.hadoop.fs.Path} where to write the classifier as a {@link
* org.apache.hadoop.io.SequenceFile}
* </ol>
*/
public static void main(String[] args) throws IOException {
DefaultOptionBuilder obuilder = new DefaultOptionBuilder();
ArgumentBuilder abuilder = new ArgumentBuilder();
GroupBuilder gbuilder = new GroupBuilder();
Option dirInputPathOpt =
obuilder
.withLongName("input")
.withRequired(true)
.withArgument(abuilder.withName("input").withMinimum(1).withMaximum(1).create())
.withDescription("The input directory path")
.withShortName("i")
.create();
Option dirOutputPathOpt =
obuilder
.withLongName("output")
.withRequired(true)
.withArgument(abuilder.withName("output").withMinimum(1).withMaximum(1).create())
.withDescription("The output directory Path")
.withShortName("o")
.create();
Option categoriesOpt =
obuilder
.withLongName("categories")
.withArgument(abuilder.withName("categories").withMinimum(1).withMaximum(1).create())
.withDescription(
"Location of the categories file. One entry per line. "
+ "Will be used to make a string match in Wikipedia Category field")
.withShortName("c")
.create();
Option exactMatchOpt =
obuilder
.withLongName("exactMatch")
.withDescription(
"If set, then the category name must exactly match the "
+ "entry in the categories file. Default is false")
.withShortName("e")
.create();
Option allOpt =
obuilder
.withLongName("all")
.withDescription("If set, Select all files. Default is false")
.withShortName("all")
.create();
Option helpOpt =
obuilder.withLongName("help").withDescription("Print out help").withShortName("h").create();
Group group =
gbuilder
.withName("Options")
.withOption(categoriesOpt)
.withOption(dirInputPathOpt)
.withOption(dirOutputPathOpt)
.withOption(exactMatchOpt)
.withOption(allOpt)
.withOption(helpOpt)
.create();
Parser parser = new Parser();
parser.setGroup(group);
parser.setHelpOption(helpOpt);
try {
CommandLine cmdLine = parser.parse(args);
if (cmdLine.hasOption(helpOpt)) {
CommandLineUtil.printHelp(group);
return;
}
String inputPath = (String) cmdLine.getValue(dirInputPathOpt);
String outputPath = (String) cmdLine.getValue(dirOutputPathOpt);
String catFile = "";
if (cmdLine.hasOption(categoriesOpt)) {
catFile = (String) cmdLine.getValue(categoriesOpt);
}
boolean all = false;
if (cmdLine.hasOption(allOpt)) {
all = true;
}
runJob(inputPath, outputPath, catFile, cmdLine.hasOption(exactMatchOpt), all);
} catch (OptionException e) {
log.error("Exception", e);
CommandLineUtil.printHelp(group);
} catch (InterruptedException e) {
log.error("Exception", e);
CommandLineUtil.printHelp(group);
} catch (ClassNotFoundException e) {
log.error("Exception", e);
CommandLineUtil.printHelp(group);
}
}
public static void main(String[] args)
throws IOException, InterruptedException, ClassNotFoundException {
DefaultOptionBuilder obuilder = new DefaultOptionBuilder();
ArgumentBuilder abuilder = new ArgumentBuilder();
GroupBuilder gbuilder = new GroupBuilder();
Option inputDirOpt = DefaultOptionCreator.inputOption().create();
Option outputOpt = DefaultOptionCreator.outputOption().create();
Option helpOpt = DefaultOptionCreator.helpOption();
Option recordSplitterOpt =
obuilder
.withLongName("splitterPattern")
.withArgument(
abuilder.withName("splitterPattern").withMinimum(1).withMaximum(1).create())
.withDescription(
"Regular Expression pattern used to split given line into fields."
+ " Default value splits comma or tab separated fields."
+ " Default Value: \"[ ,\\t]*\\t[ ,\\t]*\" ")
.withShortName("regex")
.create();
Option encodingOpt =
obuilder
.withLongName("encoding")
.withArgument(abuilder.withName("encoding").withMinimum(1).withMaximum(1).create())
.withDescription("(Optional) The file encoding. Default value: UTF-8")
.withShortName("e")
.create();
Group group =
gbuilder
.withName("Options")
.withOption(inputDirOpt)
.withOption(outputOpt)
.withOption(helpOpt)
.withOption(recordSplitterOpt)
.withOption(encodingOpt)
.create();
try {
Parser parser = new Parser();
parser.setGroup(group);
CommandLine cmdLine = parser.parse(args);
if (cmdLine.hasOption(helpOpt)) {
CommandLineUtil.printHelp(group);
return;
}
Parameters params = new Parameters();
if (cmdLine.hasOption(recordSplitterOpt)) {
params.set("splitPattern", (String) cmdLine.getValue(recordSplitterOpt));
}
String encoding = "UTF-8";
if (cmdLine.hasOption(encodingOpt)) {
encoding = (String) cmdLine.getValue(encodingOpt);
}
params.set("encoding", encoding);
String inputDir = (String) cmdLine.getValue(inputDirOpt);
String outputDir = (String) cmdLine.getValue(outputOpt);
params.set("input", inputDir);
params.set("output", outputDir);
params.set("groupingFieldCount", "2");
params.set("gfield0", "1");
params.set("gfield1", "2");
params.set("selectedFieldCount", "1");
params.set("field0", "3");
params.set("maxTransactionLength", "100");
KeyBasedStringTupleGrouper.startJob(params);
} catch (OptionException ex) {
CommandLineUtil.printHelp(group);
}
}
public static void main(String[] args) throws IOException {
DefaultOptionBuilder obuilder = new DefaultOptionBuilder();
ArgumentBuilder abuilder = new ArgumentBuilder();
GroupBuilder gbuilder = new GroupBuilder();
Option inputOpt =
obuilder
.withLongName("input")
.withRequired(true)
.withArgument(abuilder.withName("input").withMinimum(1).withMaximum(1).create())
.withDescription(
"The file or directory containing the ARFF files. If it is a directory, all .arff files will be converted")
.withShortName("d")
.create();
Option outputOpt =
obuilder
.withLongName("output")
.withRequired(true)
.withArgument(abuilder.withName("output").withMinimum(1).withMaximum(1).create())
.withDescription(
"The output directory. Files will have the same name as the input, but with the extension .mvc")
.withShortName("o")
.create();
Option maxOpt =
obuilder
.withLongName("max")
.withRequired(false)
.withArgument(abuilder.withName("max").withMinimum(1).withMaximum(1).create())
.withDescription(
"The maximum number of vectors to output. If not specified, then it will loop over all docs")
.withShortName("m")
.create();
Option dictOutOpt =
obuilder
.withLongName("dictOut")
.withRequired(true)
.withArgument(abuilder.withName("dictOut").withMinimum(1).withMaximum(1).create())
.withDescription("The file to output the label bindings")
.withShortName("t")
.create();
Option jsonDictonaryOpt =
obuilder
.withLongName("json-dictonary")
.withRequired(false)
.withDescription("Write dictonary in JSON format")
.withShortName("j")
.create();
Option delimiterOpt =
obuilder
.withLongName("delimiter")
.withRequired(false)
.withArgument(abuilder.withName("delimiter").withMinimum(1).withMaximum(1).create())
.withDescription("The delimiter for outputing the dictionary")
.withShortName("l")
.create();
Option helpOpt =
obuilder.withLongName("help").withDescription("Print out help").withShortName("h").create();
Group group =
gbuilder
.withName("Options")
.withOption(inputOpt)
.withOption(outputOpt)
.withOption(maxOpt)
.withOption(helpOpt)
.withOption(dictOutOpt)
.withOption(jsonDictonaryOpt)
.withOption(delimiterOpt)
.create();
try {
Parser parser = new Parser();
parser.setGroup(group);
CommandLine cmdLine = parser.parse(args);
if (cmdLine.hasOption(helpOpt)) {
CommandLineUtil.printHelp(group);
return;
}
if (cmdLine.hasOption(inputOpt)) { // Lucene case
File input = new File(cmdLine.getValue(inputOpt).toString());
long maxDocs = Long.MAX_VALUE;
if (cmdLine.hasOption(maxOpt)) {
maxDocs = Long.parseLong(cmdLine.getValue(maxOpt).toString());
}
if (maxDocs < 0) {
throw new IllegalArgumentException("maxDocs must be >= 0");
}
String outDir = cmdLine.getValue(outputOpt).toString();
log.info("Output Dir: {}", outDir);
String delimiter =
cmdLine.hasOption(delimiterOpt) ? cmdLine.getValue(delimiterOpt).toString() : "\t";
File dictOut = new File(cmdLine.getValue(dictOutOpt).toString());
boolean jsonDictonary = cmdLine.hasOption(jsonDictonaryOpt);
ARFFModel model = new MapBackedARFFModel();
if (input.exists() && input.isDirectory()) {
File[] files =
input.listFiles(
new FilenameFilter() {
@Override
public boolean accept(File file, String name) {
return name.endsWith(".arff");
}
});
for (File file : files) {
writeFile(outDir, file, maxDocs, model, dictOut, delimiter, jsonDictonary);
}
} else {
writeFile(outDir, input, maxDocs, model, dictOut, delimiter, jsonDictonary);
}
}
} catch (OptionException e) {
log.error("Exception", e);
CommandLineUtil.printHelp(group);
}
}
public static int main2(String[] args, Configuration conf) throws Exception {
DefaultOptionBuilder obuilder = new DefaultOptionBuilder();
ArgumentBuilder abuilder = new ArgumentBuilder();
GroupBuilder gbuilder = new GroupBuilder();
Option helpOpt = DefaultOptionCreator.helpOption();
Option inputDirOpt =
obuilder
.withLongName("input")
.withRequired(true)
.withArgument(abuilder.withName("input").withMinimum(1).withMaximum(1).create())
.withDescription(
"The Directory on HDFS containing the collapsed, properly formatted files having "
+ "one doc per line")
.withShortName("i")
.create();
Option dictOpt =
obuilder
.withLongName("dictionary")
.withRequired(false)
.withArgument(abuilder.withName("dictionary").withMinimum(1).withMaximum(1).create())
.withDescription("The path to the term-dictionary format is ... ")
.withShortName("d")
.create();
Option dfsOpt =
obuilder
.withLongName("dfs")
.withRequired(false)
.withArgument(abuilder.withName("dfs").withMinimum(1).withMaximum(1).create())
.withDescription("HDFS namenode URI")
.withShortName("dfs")
.create();
Option numTopicsOpt =
obuilder
.withLongName("numTopics")
.withRequired(true)
.withArgument(abuilder.withName("numTopics").withMinimum(1).withMaximum(1).create())
.withDescription("Number of topics to learn")
.withShortName("top")
.create();
Option outputTopicFileOpt =
obuilder
.withLongName("topicOutputFile")
.withRequired(true)
.withArgument(
abuilder.withName("topicOutputFile").withMinimum(1).withMaximum(1).create())
.withDescription("File to write out p(term | topic)")
.withShortName("to")
.create();
Option outputDocFileOpt =
obuilder
.withLongName("docOutputFile")
.withRequired(true)
.withArgument(abuilder.withName("docOutputFile").withMinimum(1).withMaximum(1).create())
.withDescription("File to write out p(topic | docid)")
.withShortName("do")
.create();
Option alphaOpt =
obuilder
.withLongName("alpha")
.withRequired(false)
.withArgument(
abuilder
.withName("alpha")
.withMinimum(1)
.withMaximum(1)
.withDefault("0.1")
.create())
.withDescription("Smoothing parameter for p(topic | document) prior")
.withShortName("a")
.create();
Option etaOpt =
obuilder
.withLongName("eta")
.withRequired(false)
.withArgument(
abuilder.withName("eta").withMinimum(1).withMaximum(1).withDefault("0.1").create())
.withDescription("Smoothing parameter for p(term | topic)")
.withShortName("e")
.create();
Option maxIterOpt =
obuilder
.withLongName("maxIterations")
.withRequired(false)
.withArgument(
abuilder
.withName("maxIterations")
.withMinimum(1)
.withMaximum(1)
.withDefault(10)
.create())
.withDescription("Maximum number of training passes")
.withShortName("m")
.create();
Option modelCorpusFractionOption =
obuilder
.withLongName("modelCorpusFraction")
.withRequired(false)
.withArgument(
abuilder
.withName("modelCorpusFraction")
.withMinimum(1)
.withMaximum(1)
.withDefault(0.0)
.create())
.withShortName("mcf")
.withDescription("For online updates, initial value of |model|/|corpus|")
.create();
Option burnInOpt =
obuilder
.withLongName("burnInIterations")
.withRequired(false)
.withArgument(
abuilder
.withName("burnInIterations")
.withMinimum(1)
.withMaximum(1)
.withDefault(5)
.create())
.withDescription("Minimum number of iterations")
.withShortName("b")
.create();
Option convergenceOpt =
obuilder
.withLongName("convergence")
.withRequired(false)
.withArgument(
abuilder
.withName("convergence")
.withMinimum(1)
.withMaximum(1)
.withDefault("0.0")
.create())
.withDescription("Fractional rate of perplexity to consider convergence")
.withShortName("c")
.create();
Option reInferDocTopicsOpt =
obuilder
.withLongName("reInferDocTopics")
.withRequired(false)
.withArgument(
abuilder
.withName("reInferDocTopics")
.withMinimum(1)
.withMaximum(1)
.withDefault("no")
.create())
.withDescription("re-infer p(topic | doc) : [no | randstart | continue]")
.withShortName("rdt")
.create();
Option numTrainThreadsOpt =
obuilder
.withLongName("numTrainThreads")
.withRequired(false)
.withArgument(
abuilder
.withName("numTrainThreads")
.withMinimum(1)
.withMaximum(1)
.withDefault("1")
.create())
.withDescription("number of threads to train with")
.withShortName("ntt")
.create();
Option numUpdateThreadsOpt =
obuilder
.withLongName("numUpdateThreads")
.withRequired(false)
.withArgument(
abuilder
.withName("numUpdateThreads")
.withMinimum(1)
.withMaximum(1)
.withDefault("1")
.create())
.withDescription("number of threads to update the model with")
.withShortName("nut")
.create();
Option verboseOpt =
obuilder
.withLongName("verbose")
.withRequired(false)
.withArgument(
abuilder
.withName("verbose")
.withMinimum(1)
.withMaximum(1)
.withDefault("false")
.create())
.withDescription(
"print verbose information, like top-terms in each topic, during iteration")
.withShortName("v")
.create();
Group group =
gbuilder
.withName("Options")
.withOption(inputDirOpt)
.withOption(numTopicsOpt)
.withOption(alphaOpt)
.withOption(etaOpt)
.withOption(maxIterOpt)
.withOption(burnInOpt)
.withOption(convergenceOpt)
.withOption(dictOpt)
.withOption(reInferDocTopicsOpt)
.withOption(outputDocFileOpt)
.withOption(outputTopicFileOpt)
.withOption(dfsOpt)
.withOption(numTrainThreadsOpt)
.withOption(numUpdateThreadsOpt)
.withOption(modelCorpusFractionOption)
.withOption(verboseOpt)
.create();
try {
Parser parser = new Parser();
parser.setGroup(group);
parser.setHelpOption(helpOpt);
CommandLine cmdLine = parser.parse(args);
if (cmdLine.hasOption(helpOpt)) {
CommandLineUtil.printHelp(group);
return -1;
}
String inputDirString = (String) cmdLine.getValue(inputDirOpt);
String dictDirString = cmdLine.hasOption(dictOpt) ? (String) cmdLine.getValue(dictOpt) : null;
int numTopics = Integer.parseInt((String) cmdLine.getValue(numTopicsOpt));
double alpha = Double.parseDouble((String) cmdLine.getValue(alphaOpt));
double eta = Double.parseDouble((String) cmdLine.getValue(etaOpt));
int maxIterations = Integer.parseInt((String) cmdLine.getValue(maxIterOpt));
int burnInIterations = (Integer) cmdLine.getValue(burnInOpt);
double minFractionalErrorChange =
Double.parseDouble((String) cmdLine.getValue(convergenceOpt));
int numTrainThreads = Integer.parseInt((String) cmdLine.getValue(numTrainThreadsOpt));
int numUpdateThreads = Integer.parseInt((String) cmdLine.getValue(numUpdateThreadsOpt));
String topicOutFile = (String) cmdLine.getValue(outputTopicFileOpt);
String docOutFile = (String) cmdLine.getValue(outputDocFileOpt);
// String reInferDocTopics = (String)cmdLine.getValue(reInferDocTopicsOpt);
boolean verbose = Boolean.parseBoolean((String) cmdLine.getValue(verboseOpt));
double modelCorpusFraction = (Double) cmdLine.getValue(modelCorpusFractionOption);
long start = System.nanoTime();
if (conf.get("fs.default.name") == null) {
String dfsNameNode = (String) cmdLine.getValue(dfsOpt);
conf.set("fs.default.name", dfsNameNode);
}
String[] terms = loadDictionary(dictDirString, conf);
logTime("dictionary loading", System.nanoTime() - start);
start = System.nanoTime();
Matrix corpus = loadVectors(inputDirString, conf);
logTime("vector seqfile corpus loading", System.nanoTime() - start);
start = System.nanoTime();
InMemoryCollapsedVariationalBayes0 cvb0 =
new InMemoryCollapsedVariationalBayes0(
corpus,
terms,
numTopics,
alpha,
eta,
numTrainThreads,
numUpdateThreads,
modelCorpusFraction);
logTime("cvb0 init", System.nanoTime() - start);
start = System.nanoTime();
cvb0.setVerbose(verbose);
cvb0.iterateUntilConvergence(minFractionalErrorChange, maxIterations, burnInIterations);
logTime("total training time", System.nanoTime() - start);
/*
if ("randstart".equalsIgnoreCase(reInferDocTopics)) {
cvb0.inferDocuments(0.0, 100, true);
} else if ("continue".equalsIgnoreCase(reInferDocTopics)) {
cvb0.inferDocuments(0.0, 100, false);
}
*/
start = System.nanoTime();
cvb0.writeModel(new Path(topicOutFile));
DistributedRowMatrixWriter.write(new Path(docOutFile), conf, cvb0.docTopicCounts);
logTime("printTopics", System.nanoTime() - start);
} catch (OptionException e) {
log.error("Error while parsing options", e);
CommandLineUtil.printHelp(group);
}
return 0;
}
/**
* Takes in two arguments:
*
* <ol>
* <li>The input {@link org.apache.hadoop.fs.Path} where the input documents live
* <li>The output {@link org.apache.hadoop.fs.Path} where to write the classifier as a {@link
* org.apache.hadoop.io.SequenceFile}
* </ol>
*/
public static void main(String[] args) throws IOException, InterruptedException {
DefaultOptionBuilder obuilder = new DefaultOptionBuilder();
ArgumentBuilder abuilder = new ArgumentBuilder();
GroupBuilder gbuilder = new GroupBuilder();
Option dirInputPathOpt = DefaultOptionCreator.inputOption().create();
Option dirOutputPathOpt = DefaultOptionCreator.outputOption().create();
Option categoriesOpt =
obuilder
.withLongName("categories")
.withRequired(true)
.withArgument(abuilder.withName("categories").withMinimum(1).withMaximum(1).create())
.withDescription(
"Location of the categories file. One entry per line. "
+ "Will be used to make a string match in Wikipedia Category field")
.withShortName("c")
.create();
Option exactMatchOpt =
obuilder
.withLongName("exactMatch")
.withDescription(
"If set, then the category name must exactly match the "
+ "entry in the categories file. Default is false")
.withShortName("e")
.create();
Option analyzerOpt =
obuilder
.withLongName("analyzer")
.withRequired(false)
.withArgument(abuilder.withName("analyzer").withMinimum(1).withMaximum(1).create())
.withDescription("The analyzer to use, must have a no argument constructor")
.withShortName("a")
.create();
Option helpOpt = DefaultOptionCreator.helpOption();
Group group =
gbuilder
.withName("Options")
.withOption(categoriesOpt)
.withOption(dirInputPathOpt)
.withOption(dirOutputPathOpt)
.withOption(exactMatchOpt)
.withOption(analyzerOpt)
.withOption(helpOpt)
.create();
Parser parser = new Parser();
parser.setGroup(group);
try {
CommandLine cmdLine = parser.parse(args);
if (cmdLine.hasOption(helpOpt)) {
CommandLineUtil.printHelp(group);
return;
}
String inputPath = (String) cmdLine.getValue(dirInputPathOpt);
String outputPath = (String) cmdLine.getValue(dirOutputPathOpt);
String catFile = (String) cmdLine.getValue(categoriesOpt);
Class<? extends Analyzer> analyzerClass = WikipediaAnalyzer.class;
if (cmdLine.hasOption(analyzerOpt)) {
String className = cmdLine.getValue(analyzerOpt).toString();
analyzerClass = Class.forName(className).asSubclass(Analyzer.class);
// try instantiating it, b/c there isn't any point in setting it if
// you can't instantiate it
ClassUtils.instantiateAs(analyzerClass, Analyzer.class);
}
runJob(inputPath, outputPath, catFile, cmdLine.hasOption(exactMatchOpt), analyzerClass);
} catch (OptionException e) {
log.error("Exception", e);
CommandLineUtil.printHelp(group);
} catch (ClassNotFoundException e) {
log.error("Exception", e);
CommandLineUtil.printHelp(group);
}
}
public static void main(String[] args) throws Exception {
DefaultOptionBuilder obuilder = new DefaultOptionBuilder();
ArgumentBuilder abuilder = new ArgumentBuilder();
GroupBuilder gbuilder = new GroupBuilder();
Option inputOpt =
obuilder
.withLongName("input")
.withRequired(true)
.withArgument(abuilder.withName("input").withMinimum(1).withMaximum(1).create())
.withDescription("Path to an LDA output (a state)")
.withShortName("i")
.create();
Option dictOpt =
obuilder
.withLongName("dict")
.withRequired(true)
.withArgument(abuilder.withName("dict").withMinimum(1).withMaximum(1).create())
.withDescription(
"Dictionary to read in, in the same format as one created by "
+ "org.apache.mahout.utils.vectors.lucene.Driver")
.withShortName("d")
.create();
Option outOpt =
obuilder
.withLongName("output")
.withRequired(false)
.withArgument(abuilder.withName("output").withMinimum(1).withMaximum(1).create())
.withDescription("Output directory to write top words")
.withShortName("o")
.create();
Option wordOpt =
obuilder
.withLongName("words")
.withRequired(false)
.withArgument(
abuilder.withName("words").withMinimum(0).withMaximum(1).withDefault("20").create())
.withDescription("Number of words to print")
.withShortName("w")
.create();
Option dictTypeOpt =
obuilder
.withLongName("dictionaryType")
.withRequired(false)
.withArgument(
abuilder.withName("dictionaryType").withMinimum(1).withMaximum(1).create())
.withDescription("The dictionary file type (text|sequencefile)")
.withShortName("dt")
.create();
Option helpOpt =
obuilder.withLongName("help").withDescription("Print out help").withShortName("h").create();
Group group =
gbuilder
.withName("Options")
.withOption(dictOpt)
.withOption(outOpt)
.withOption(wordOpt)
.withOption(inputOpt)
.withOption(dictTypeOpt)
.create();
try {
Parser parser = new Parser();
parser.setGroup(group);
CommandLine cmdLine = parser.parse(args);
if (cmdLine.hasOption(helpOpt)) {
CommandLineUtil.printHelp(group);
return;
}
String input = cmdLine.getValue(inputOpt).toString();
String dictFile = cmdLine.getValue(dictOpt).toString();
int numWords = 20;
if (cmdLine.hasOption(wordOpt)) {
numWords = Integer.parseInt(cmdLine.getValue(wordOpt).toString());
}
Configuration config = new Configuration();
String dictionaryType = "text";
if (cmdLine.hasOption(dictTypeOpt)) {
dictionaryType = cmdLine.getValue(dictTypeOpt).toString();
}
List<String> wordList;
if ("text".equals(dictionaryType)) {
wordList = Arrays.asList(VectorHelper.loadTermDictionary(new File(dictFile)));
} else if ("sequencefile".equals(dictionaryType)) {
wordList = Arrays.asList(VectorHelper.loadTermDictionary(config, dictFile));
} else {
throw new IllegalArgumentException("Invalid dictionary format");
}
List<Queue<Pair<String, Double>>> topWords =
topWordsForTopics(input, config, wordList, numWords);
File output = null;
if (cmdLine.hasOption(outOpt)) {
output = new File(cmdLine.getValue(outOpt).toString());
if (!output.exists() && !output.mkdirs()) {
throw new IOException("Could not create directory: " + output);
}
}
printTopWords(topWords, output);
} catch (OptionException e) {
CommandLineUtil.printHelp(group);
throw e;
}
}