Added the nbTrain option to train the same model many times with different seeds

MLP/include/MLP.hpp

@@ -37,6 +37,8 @@ class MLP
   /// @brief The seed that will be used by RNG (srand and dynet)
   static int randomSeed;
 
+  static bool dynetIsInit;
+
   /// @brief Get the string corresponding to an Activation.
   ///
   /// @param a The activation.

MLP/src/MLP.cpp

@@ -7,6 +7,7 @@
 #include <dynet/io.h>
 
 int MLP::randomSeed = 0;
+bool MLP::dynetIsInit = false;
 
 std::string MLP::activation2str(Activation a)
 {
@@ -72,12 +73,10 @@ MLP::Activation MLP::str2activation(std::string s)
 
 void MLP::initDynet()
 {
-  static bool init = false;
-
-  if(init)
+  if(dynetIsInit)
     return;
 
-  init = true;
+  dynetIsInit = true;
   dynet::initialize(getDefaultParams());
 }
 

maca_common/include/Dict.hpp

@@ -203,6 +203,8 @@ class Dict
   /// @param directory The directory in which we will save every Dict.
   /// @param namePrefix The prefix of the name of the dicts we need to save.
   static void saveDicts(const std::string & directory, const std::string & namePrefix);
+  /// @brief Delete all Dicts.
+  static void deleteDicts();
   /// @brief Save the current Dict in the corresponding file.
   void save();
   /// @brief Get the vector value of an entry.

maca_common/include/ProgramParameters.hpp

@@ -12,6 +12,7 @@ struct ProgramParameters
   static std::string input;
   static std::string expName;
   static std::string expPath;
+  static std::string baseExpName;
   static std::string langPath;
   static std::string templatePath;
   static std::string templateName;
@@ -41,6 +42,7 @@ struct ProgramParameters
   static float dynamicProbability;
   static bool showFeatureRepresentation;
   static int iterationSize;
+  static int nbTrain;
 
   private :
 

maca_common/src/Dict.cpp

@@ -416,3 +416,8 @@ void Dict::printForDebug(FILE * output)
   fprintf(output, "Dict name \'%s\' nbElems = %lu\n", name.c_str(), str2vec.size());
 }
 
+void Dict::deleteDicts()
+{
+  str2dict.clear();
+}
+

maca_common/src/ProgramParameters.cpp

@@ -6,6 +6,7 @@ ProgramParameters::ProgramParameters()
 
 std::string ProgramParameters::input;
 std::string ProgramParameters::expName;
+std::string ProgramParameters::baseExpName;
 std::string ProgramParameters::expPath;
 std::string ProgramParameters::langPath;
 std::string ProgramParameters::templatePath;
@@ -36,3 +37,4 @@ int ProgramParameters::dynamicEpoch;
 float ProgramParameters::dynamicProbability;
 bool ProgramParameters::showFeatureRepresentation;
 int ProgramParameters::iterationSize;
+int ProgramParameters::nbTrain;

trainer/src/macaon_train.cpp

@@ -54,6 +54,8 @@ po::options_description getOptionsDescription()
       "Learning rate of the optimizer")
     ("seed,s", po::value<int>()->default_value(100),
       "The random seed that will initialize RNG")
+    ("nbTrain", po::value<int>()->default_value(0),
+      "The number of models that will be trained, with only the random seed changing")
     ("duplicates", po::value<bool>()->default_value(true),
       "Remove identical training examples")
     ("showFeatureRepresentation", po::value<bool>()->default_value(false),
@@ -209,6 +211,7 @@ int main(int argc, char * argv[])
   po::variables_map vm = checkOptions(od, argc, argv);
 
   ProgramParameters::expName = vm["expName"].as<std::string>();
+  ProgramParameters::baseExpName = ProgramParameters::expName;
   ProgramParameters::templateName = vm["templateName"].as<std::string>();
   ProgramParameters::tmName = vm["tm"].as<std::string>();
   ProgramParameters::bdName = vm["bd"].as<std::string>();
@@ -219,6 +222,7 @@ int main(int argc, char * argv[])
   ProgramParameters::lang = vm["lang"].as<std::string>();
   ProgramParameters::nbIter = vm["nbiter"].as<int>();
   ProgramParameters::seed = vm["seed"].as<int>();
+  ProgramParameters::nbTrain = vm["nbTrain"].as<int>();
   ProgramParameters::removeDuplicates = vm["duplicates"].as<bool>();
   ProgramParameters::interactive = vm["interactive"].as<bool>();
   ProgramParameters::shuffleExamples = vm["shuffle"].as<bool>();
@@ -232,11 +236,23 @@ int main(int argc, char * argv[])
   ProgramParameters::showFeatureRepresentation = vm["showFeatureRepresentation"].as<bool>();
   ProgramParameters::iterationSize = vm["iterationSize"].as<int>();
 
-  for (int i = 0; i < 10; i++)
+  if (ProgramParameters::nbTrain)
+  {
+    for (int i = 0; i < ProgramParameters::nbTrain; i++)
+    {
+      fprintf(stderr, "Training number %d / %d :\n", i+1, ProgramParameters::nbTrain);
+      ProgramParameters::expName = ProgramParameters::baseExpName + "_" + std::to_string(i);
+      updatePaths();
+      createExpPath();
+      Dict::deleteDicts();
+      launchTraining();
+    }
+  }
+  else
   {
-    ProgramParameters::expName += "_" + std::to_string(i);
     updatePaths();
     createExpPath();
+    Dict::deleteDicts();
     launchTraining();
   }