internal static FileInfo PreprocessLanguage(
FileInfo languageFile,
DirectoryInfo directory,
string languageCode,
FileInfo spmModel,
bool includePlaceholderTags,
bool includeTagPairs)
{
var preprocessedFile = new FileInfo(Path.Combine(directory.FullName, $"preprocessed_{languageFile.Name}"));
//Marian doesn't like spaces in names
var spFile = new FileInfo(Path.Combine(directory.FullName, $"sp_{languageFile.Name.Replace(" ", "_")}"));
using (var rawFile = languageFile.OpenText())
using (var preprocessedWriter = new StreamWriter(preprocessedFile.FullName))
{
String line;
while ((line = rawFile.ReadLine()) != null)
{
var preprocessedLine = MarianHelper.PreprocessLine(line, languageCode, includePlaceholderTags, includeTagPairs);
preprocessedWriter.WriteLine(preprocessedLine);
}
}
var spArgs = $"\"{preprocessedFile.FullName}\" --model \"{spmModel.FullName}\" --output \"{spFile.FullName}\"";
var spmProcess = MarianHelper.StartProcessInBackgroundWithRedirects("Preprocessing\\spm_encode.exe", spArgs);
spmProcess.WaitForExit();
return(spFile);
}
//Callback can be used to do different things with translation output/input (default is to save in translation cache)
internal Process BatchTranslate(
IEnumerable <string> input,
FileInfo spOutput,
Boolean preprocessedInput = false,
Boolean storeTranslations = false)
{
if (storeTranslations)
{
this.OutputReady += (x, y) => this.WriteToTranslationDb(x, y, input, spOutput);
}
Log.Information($"Starting batch translator for model {this.SystemName}.");
var cmd = "TranslateBatchSentencePiece.bat";
FileInfo spInput = this.PreprocessInput(input, preprocessedInput);
//TODO: check the translation cache for translations beforehand, and only translate new
//segments (also change translation cache to account for different decoder configs for
//same systems, i.e. keep track of decoder settings)
FileInfo transAndAlign = new FileInfo($"{spOutput.FullName}.transandalign");
var args = $"{this.modelDir.FullName} {spInput.FullName} {transAndAlign.FullName} --log-level=info --quiet";
EventHandler exitHandler = (x, y) => BatchProcess_Exited(transAndAlign, spOutput, x, y);
var batchProcess = MarianHelper.StartProcessInBackgroundWithRedirects(cmd, args, exitHandler);
return(batchProcess);
}
internal Process BatchTranslate(
IEnumerable <string> input,
FileInfo spOutput,
Boolean preprocessedInput = false,
Boolean storeTranslations = false)
{
Log.Information($"Starting batch translator for model {this.SystemName}.");
var srcFile = MarianHelper.LinesToFile(input, this.SourceCode);
FileInfo spInput;
if (!preprocessedInput)
{
FileInfo sourceSegModel =
this.modelDir.GetFiles().Where(x => Regex.IsMatch(x.Name, "source.(spm|bpe)")).Single();
spInput = MarianHelper.PreprocessLanguage(
srcFile,
new DirectoryInfo(Path.GetTempPath()),
this.TargetCode,
sourceSegModel,
this.includePlaceholderTags,
this.includeTagPairs);
}
else
{
spInput = srcFile;
}
//TODO: check the translation cache for translations beforehand, and only translate new
//segments (also change translation cache to account for different decoder configs for
//same systems, i.e. keep track of decoder settings)
FileInfo transAndAlign = new FileInfo($"{spOutput.FullName}.transandalign");
var args = $"\"{this.modelDir.FullName}\" \"{spInput.FullName}\" \"{transAndAlign.FullName}\" --log-level=info --quiet";
if (storeTranslations)
{
this.OutputReady += (x, y) => this.WriteToTranslationDb(x, y, input, spInput, transAndAlign);
}
EventHandler exitHandler = (x, y) => BatchProcess_Exited(transAndAlign, spOutput, x, y);
var cmd = "TranslateBatchSentencePiece.bat";
var batchProcess = MarianHelper.StartProcessInBackgroundWithRedirects(cmd, args, exitHandler);
return(batchProcess);
}
internal static void GenerateAlignments(FileInfo spSource, FileInfo spTarget, FileInfo alignmentFile, FileInfo priorsFile)
{
var alignArgs = $"-s \"{spSource.FullName}\" -t \"{spTarget.FullName}\" -f \"{alignmentFile.FullName}.fwd\" -r \"{alignmentFile.FullName}.rev\"";
Log.Information($"Aligning fine-tuning corpus with args {alignArgs}");
var alignProcess = MarianHelper.StartProcessInBackgroundWithRedirects("python Alignment\\align.py", alignArgs);
alignProcess.WaitForExit();
var symmetryArgs = $"-c grow-diag-final -i \"{alignmentFile.FullName}.fwd\" -j \"{alignmentFile.FullName}.rev\" > \"{alignmentFile.FullName}\"";
Log.Information($"Symmetrisizing alignment with args {symmetryArgs}");
var symmetryProcess = MarianHelper.StartProcessInBackgroundWithRedirects("Alignment\\atools.exe", symmetryArgs);
symmetryProcess.WaitForExit();
}
private Process StartTraining()
{
var configPath = Path.Combine(this.customDir.FullName, OpusCatMTEngineSettings.Default.CustomizationBaseConfig);
var deserializer = new Deserializer();
MarianTrainerConfig trainingConfig;
using (var reader = new StreamReader(configPath))
{
trainingConfig = deserializer.Deserialize <MarianTrainerConfig>(reader);
}
this.trainingLog.TrainingConfig = trainingConfig;
//var trainingArgs = $"--config {configPath} --log-level=warn";
var trainingArgs = $"--config \"{configPath}\" --log-level=info"; // --quiet";
var trainProcess = MarianHelper.StartProcessInBackgroundWithRedirects(
Path.Combine(OpusCatMTEngineSettings.Default.MarianDir, "marian.exe"), trainingArgs, this.MarianExitHandler, this.MarianProgressHandler);
return(trainProcess);
}
internal static FileInfo PreprocessLanguage(
FileInfo languageFile,
DirectoryInfo directory,
string languageCode,
FileInfo segmentationModel,
bool includePlaceholderTags,
bool includeTagPairs,
string targetLanguageToPrefix = null)
{
var preprocessedFile = new FileInfo(Path.Combine(directory.FullName, $"preprocessed_{languageFile.Name}"));
using (var rawFile = languageFile.OpenText())
using (var preprocessedWriter = new StreamWriter(preprocessedFile.FullName))
{
String line;
while ((line = rawFile.ReadLine()) != null)
{
var preprocessedLine = MarianHelper.PreprocessLine(line, languageCode, includePlaceholderTags, includeTagPairs);
preprocessedWriter.WriteLine(preprocessedLine);
}
}
//Marian doesn't like spaces in names
var segmentedFile = new FileInfo(Path.Combine(directory.FullName, $"seg_{languageFile.Name.Replace(" ", "_")}"));
switch (segmentationModel.Extension)
{
case ".spm":
var spArgs = $"\"{preprocessedFile.FullName}\" --model \"{segmentationModel.FullName}\" --output \"{segmentedFile.FullName}\"";
var segmentationProcess = MarianHelper.StartProcessInBackgroundWithRedirects("Preprocessing\\spm_encode.exe", spArgs);
segmentationProcess.WaitForExit();
break;
case ".bpe":
//Truecasing is not used in any models, so this is a dummy tc model (empty). So it does not
//matter is source.tcmodel is used for target language.
var tcModelPath = $@"{directory.FullName}\source.tcmodel";
var mosesProcess = MarianHelper.StartProcessInBackgroundWithRedirects(
$"type {preprocessedFile.FullName} | Preprocessing\\StartMosesBpePreprocessPipe.bat {languageCode} \"{tcModelPath}\" \"{segmentationModel.FullName}\" > {segmentedFile.FullName}");
mosesProcess.WaitForExit();
break;
default:
segmentationProcess = null;
throw new Exception("No segmentation model found");
break;
}
if (targetLanguageToPrefix != null)
{
var segmentedWithTargetPrefix = new FileInfo(Path.Combine(directory.FullName, $"prefix_{languageFile.Name.Replace(" ", "_")}"));
using (var segFile = segmentedFile.OpenText())
using (var prefixWriter = new StreamWriter(segmentedWithTargetPrefix.FullName))
{
String line;
while ((line = segFile.ReadLine()) != null)
{
var prefixedLine = $">>{targetLanguageToPrefix}<< {line}";
prefixWriter.WriteLine(prefixedLine);
}
}
return(segmentedWithTargetPrefix);
}
else
{
return(segmentedFile);
}
}
