model changes

piano_transcription/train_model.py

@@ -113,17 +113,19 @@ def run(model, model_name, learn_rate, batch_size, split, k_samples):
         valid_loss[epoch] = valid_loss_sum / valid_batches
 
         # Then we print the results for this epoch:
-        print("Epoch %3d of %d took %1.3f s (valid: %1.3f s) -- patience: %d" %
+        print("\rEpoch %3d of %d took %1.3f s (valid: %1.3f s) -- patience: %d                                         " %
               (epoch + 1, MAX_NUMEPOCHS, time.time() - start_time, time.time() - valid_start_time, cur_patience))
         error = train_loss_sum / train_batches
         print("  training loss: %1.3f valid loss: %1.3f" % (error, valid_loss[epoch]))
-        if epoch == 0 or valid_loss[epoch] < valid_loss[best_valid_loss_epoch]:
+        better = valid_loss[epoch] < valid_loss[best_valid_loss_epoch]
+        if epoch == 0 or better:
             best_valid_loss_epoch = epoch
+            np.savez(os.path.join(out_directory, 'best_model.npz'), *lasagne.layers.get_all_param_values(network))
             print('  new best validation loss at epoch %3d: %1.3f' % (epoch, valid_loss[epoch]))
 
         np.save(os.path.join(out_directory, 'losses.npy'), [train_loss[:epoch], valid_loss[:epoch]])
 
-        if epoch > 0 and valid_loss[epoch] >= valid_loss[best_valid_loss_epoch]:
+        if epoch > 0 and not better:
             cur_patience -= 1
             if cur_patience <= 0:
                 if refinements > 0:
@@ -139,7 +141,7 @@ def run(model, model_name, learn_rate, batch_size, split, k_samples):
             cur_patience = patience
 
     # Optionally, you could now dump the network weights to a file like this:
-    np.savez(os.path.join(out_directory, 'best_model.npz'), *lasagne.layers.get_all_param_values(network))
+
     #
     # And load them again later on like this:
     # with np.load('model.npz') as f: