diff --git a/parser/lstm-parse.cc b/parser/lstm-parse.cc
index b96732e..a3816ce 100644
--- a/parser/lstm-parse.cc
+++ b/parser/lstm-parse.cc
@@ -71,7 +71,7 @@ void InitCommandLine(int argc, char** argv, po::variables_map* conf) {
         ("lstm_input_dim", po::value<unsigned>()->default_value(60), "LSTM input dimension")
         ("train,t", "Should training be run?")
         ("maxit,M", po::value<unsigned>()->default_value(8000), "Maximum number of training iterations")
-        ("tolerance", po::value<double>()->default_value(0.0), "Tolerance on dev uas for stopping training")
+        ("tolerance", po::value<double>()->default_value(-1.0), "Tolerance on dev uas for stopping training")
         ("words,w", po::value<string>(), "Pretrained word embeddings")
         ("help,h", "Help");
   po::options_description dcmdline_options;
@@ -546,7 +546,9 @@ int main(int argc, char** argv) {
   const unsigned maxit = conf["maxit"].as<unsigned>();
   cerr << "Maximum number of iterations: " << maxit << "\n";
   const double tolerance = conf["tolerance"].as<double>();
-  cerr << "Optimization tolerance: " << tolerance << "\n";
+  if (tolerance > 0.0) {
+    cerr << "Optimization tolerance: " << tolerance << "\n";
+  }
   ostringstream os;
   os << "parser_" << (USE_POS ? "pos" : "nopos")
      << '_' << LAYERS
@@ -634,7 +636,7 @@ int main(int argc, char** argv) {
     time_t time_start = std::chrono::system_clock::to_time_t(std::chrono::system_clock::now());
     cerr << "TRAINING STARTED AT: " << put_time(localtime(&time_start), "%c %Z") << endl;
     while(!requested_stop && iter < maxit &&
-        (uas < 0 || prev_uas < 0 || abs(prev_uas - uas) > tolerance)) {
+        (tolerance < 0 || uas < 0 || prev_uas < 0 || abs(prev_uas - uas) > tolerance)) {
       for (unsigned sii = 0; sii < status_every_i_iterations; ++sii) {
            if (si == corpus.nsentences) {
              si = 0;