private static List <IMetric> CreateMetrics()
{
IMetric seqGenMetric = null;
if (opts.SeqGenerationMetric.Equals("BLEU", StringComparison.InvariantCultureIgnoreCase))
{
seqGenMetric = new BleuMetric();
}
else
{
seqGenMetric = new RougeMetric();
}
List <IMetric> metrics = new List <IMetric>
{
seqGenMetric,
new LengthRatioMetric()
};
return(metrics);
}
private static void Main(string[] args)
{
try
{
//Parse command line
ArgParser argParser = new ArgParser(args, opts);
if (!opts.ConfigFilePath.IsNullOrEmpty())
{
Console.WriteLine($"Loading config file from '{opts.ConfigFilePath}'");
opts = JsonConvert.DeserializeObject <Seq2SeqClassificationOptions>(File.ReadAllText(opts.ConfigFilePath));
}
Logger.LogFile = $"{nameof(Seq2SeqClassificationConsole)}_{opts.Task}_{Utils.GetTimeStamp(DateTime.Now)}.log";
ShowOptions(args, opts);
DecodingOptions decodingOptions = opts.CreateDecodingOptions();
Seq2SeqClassification ss = null;
if (opts.Task == ModeEnums.Train)
{
// Load train corpus
var trainCorpus = new Seq2SeqClassificationCorpus(corpusFilePath: opts.TrainCorpusPath, srcLangName: opts.SrcLang, tgtLangName: opts.TgtLang, batchSize: opts.BatchSize, shuffleBlockSize: opts.ShuffleBlockSize,
maxSrcSentLength: opts.MaxTrainSrcSentLength, maxTgtSentLength: opts.MaxTrainTgtSentLength, shuffleEnums: opts.ShuffleType, tooLongSequence: opts.TooLongSequence);
// Load valid corpus
var validCorpusList = new List <Seq2SeqClassificationCorpus>();
if (!opts.ValidCorpusPaths.IsNullOrEmpty())
{
string[] validCorpusPathList = opts.ValidCorpusPaths.Split(';');
foreach (var validCorpusPath in validCorpusPathList)
{
validCorpusList.Add(new Seq2SeqClassificationCorpus(validCorpusPath, opts.SrcLang, opts.TgtLang, opts.ValBatchSize, opts.ShuffleBlockSize, opts.MaxTestSrcSentLength, opts.MaxTestTgtSentLength, shuffleEnums: opts.ShuffleType, tooLongSequence: opts.TooLongSequence));
}
}
// Create learning rate
ILearningRate learningRate = new DecayLearningRate(opts.StartLearningRate, opts.WarmUpSteps, opts.WeightsUpdateCount);
// Create optimizer
IOptimizer optimizer = Misc.CreateOptimizer(opts);
// Create metrics
IMetric seqGenMetric = null;
if (opts.SeqGenerationMetric.Equals("BLEU", StringComparison.InvariantCultureIgnoreCase))
{
seqGenMetric = new BleuMetric();
}
else
{
seqGenMetric = new RougeMetric();
}
Dictionary <int, List <IMetric> > taskId2metrics = new Dictionary <int, List <IMetric> >();
List <IMetric> task1Metrics = new List <IMetric>
{
seqGenMetric,
new LengthRatioMetric()
};
taskId2metrics.Add(1, task1Metrics);
var task0Metrics = new List <IMetric>();
if (!opts.ModelFilePath.IsNullOrEmpty() && File.Exists(opts.ModelFilePath))
{
//Incremental training
Logger.WriteLine($"Loading model from '{opts.ModelFilePath}'...");
ss = new Seq2SeqClassification(opts);
task0Metrics.Add(new MultiLabelsFscoreMetric("", ss.ClsVocab.GetAllTokens(keepBuildInTokens: false)));
}
else
{
// Load or build vocabulary
Vocab srcVocab = null;
Vocab tgtVocab = null;
Vocab clsVocab = null;
if (!opts.SrcVocab.IsNullOrEmpty() && !opts.TgtVocab.IsNullOrEmpty() && !opts.ClsVocab.IsNullOrEmpty())
{
Logger.WriteLine($"Loading source vocabulary from '{opts.SrcVocab}' and target vocabulary from '{opts.TgtVocab}' and classification vocabulary from '{opts.ClsVocab}'. Shared vocabulary is '{opts.SharedEmbeddings}'");
if (opts.SharedEmbeddings == true && (opts.SrcVocab != opts.TgtVocab))
{
throw new ArgumentException("The source and target vocabularies must be identical if their embeddings are shared.");
}
// Vocabulary files are specified, so we load them
srcVocab = new Vocab(opts.SrcVocab);
tgtVocab = new Vocab(opts.TgtVocab);
clsVocab = new Vocab(opts.ClsVocab);
}
else
{
Logger.WriteLine($"Building vocabulary from training corpus. Shared vocabulary is '{opts.SharedEmbeddings}'");
if (!opts.SrcVocab.IsNullOrEmpty())
{
Logger.WriteLine(Logger.Level.warn, ConsoleColor.Yellow, $"Source vocabulary '{opts.SrcVocab}' is not empty, but we will build it from training corpus.");
}
if (!opts.TgtVocab.IsNullOrEmpty())
{
Logger.WriteLine(Logger.Level.warn, ConsoleColor.Yellow, $"Target vocabulary '{opts.TgtVocab}' is not empty, but we will build it from training corpus.");
}
if (!opts.ClsVocab.IsNullOrEmpty())
{
Logger.WriteLine(Logger.Level.warn, ConsoleColor.Yellow, $"Classification vocabulary '{opts.ClsVocab}' is not empty, but we will build it from training corpus.");
}
// We don't specify vocabulary, so we build it from train corpus
(srcVocab, tgtVocab, clsVocab) = trainCorpus.BuildVocabs(opts.SrcVocabSize, opts.TgtVocabSize, opts.SharedEmbeddings);
}
//New training
ss = new Seq2SeqClassification(opts, srcVocab, tgtVocab, clsVocab);
task0Metrics.Add(new MultiLabelsFscoreMetric("", clsVocab.GetAllTokens(keepBuildInTokens: false)));
}
taskId2metrics.Add(0, task0Metrics);
// Add event handler for monitoring
ss.StatusUpdateWatcher += Misc.Ss_StatusUpdateWatcher;
ss.EvaluationWatcher += Ss_EvaluationWatcher;
// Kick off training
ss.Train(maxTrainingEpoch: opts.MaxEpochNum, trainCorpus: trainCorpus, validCorpusList: validCorpusList.ToArray(), learningRate: learningRate, optimizer: optimizer, taskId2metrics: taskId2metrics, decodingOptions: decodingOptions);
}
else if (opts.Task == ModeEnums.Valid)
{
Logger.WriteLine($"Evaluate model '{opts.ModelFilePath}' by valid corpus '{opts.ValidCorpusPaths}'");
// Load valid corpus
Seq2SeqClassificationCorpus validCorpus = new Seq2SeqClassificationCorpus(opts.ValidCorpusPaths, opts.SrcLang, opts.TgtLang, opts.ValBatchSize, opts.ShuffleBlockSize, opts.MaxTestSrcSentLength, opts.MaxTestTgtSentLength, shuffleEnums: opts.ShuffleType, tooLongSequence: opts.TooLongSequence);
ss = new Seq2SeqClassification(opts);
ss.EvaluationWatcher += Ss_EvaluationWatcher;
// Create metrics
IMetric seqGenMetric = null;
if (opts.SeqGenerationMetric.Equals("BLEU", StringComparison.InvariantCultureIgnoreCase))
{
seqGenMetric = new BleuMetric();
}
else
{
seqGenMetric = new RougeMetric();
}
Dictionary <int, List <IMetric> > taskId2metrics = new Dictionary <int, List <IMetric> >();
List <IMetric> task1Metrics = new List <IMetric>
{
seqGenMetric,
new LengthRatioMetric()
};
taskId2metrics.Add(1, task1Metrics);
List <IMetric> task0Metrics = new List <IMetric>()
{
new MultiLabelsFscoreMetric("", ss.ClsVocab.GetAllTokens(keepBuildInTokens: false))
};
taskId2metrics.Add(0, task0Metrics);
ss.Valid(validCorpus: validCorpus, taskId2metrics, decodingOptions: decodingOptions);
}
else if (opts.Task == ModeEnums.Test)
{
if (File.Exists(opts.OutputFile))
{
Logger.WriteLine(Logger.Level.err, ConsoleColor.Yellow, $"Output file '{opts.OutputFile}' exist. Delete it.");
File.Delete(opts.OutputFile);
}
//Test trained model
ss = new Seq2SeqClassification(opts);
Stopwatch stopwatch = Stopwatch.StartNew();
ss.Test <Seq2SeqClassificationCorpusBatch>(opts.InputTestFile, opts.OutputFile, opts.BatchSize, decodingOptions, opts.SrcSentencePieceModelPath, opts.TgtSentencePieceModelPath);
stopwatch.Stop();
Logger.WriteLine($"Test mode execution time elapsed: '{stopwatch.Elapsed}'");
}
else if (opts.Task == ModeEnums.DumpVocab)
{
ss = new Seq2SeqClassification(opts);
ss.DumpVocabToFiles(opts.SrcVocab, opts.TgtVocab, opts.ClsVocab);
}
else
{
Logger.WriteLine(Logger.Level.err, ConsoleColor.Red, $"Task '{opts.Task}' is not supported.");
argParser.Usage();
}
}
catch (Exception err)
{
Logger.WriteLine($"Exception: '{err.Message}'");
Logger.WriteLine($"Call stack: '{err.StackTrace}'");
}
}
