Add early stop after 1st epoch

cifar10-cnn.py

@@ -37,7 +37,7 @@ def main(argumentList):
 
     batch_size = int(getValue(argumentsDict, '-b', '--batch_size', 32))
 
-    epochs = int(argumentsDict.get('-e', argumentsDict.get('--epochs', 100)))
+    epochs = int(argumentsDict.get('-e', argumentsDict.get('--epochs', 10)))
 
     data_augmentation = bool(argumentsDict.get('-a', argumentsDict.get('--augmentation', True)))
 
@@ -71,6 +71,9 @@ def main(argumentList):
     conv_map = conv_map.split(',')
     full_map = full_map.split(',')
 
+    # If accuracy after 1t epoch is below this limit then break the training
+    acc_break_limit = -0.15
+
     # The data, shuffled and split between train and test sets:
     (x_train, y_train), (x_test, y_test) = cifar10.load_data()
     print('x_train shape:', x_train.shape)
@@ -132,9 +135,24 @@ def main(argumentList):
             print('Not using data augmentation.')
             model.fit(x_train, y_train,
                       batch_size=batch_size,
-                      epochs=epochs,
+                      epochs=1,
                       validation_data=(x_test, y_test),
                       shuffle=True)
+
+            # Score trained model.
+            scores = model.evaluate(x_test, y_test)
+
+            if scores[1] > acc_break_limit:
+                # If the model looks promising....
+                model.fit(x_train, y_train,
+                          batch_size=batch_size,
+                          epochs=epochs,
+                          validation_data=(x_test, y_test),
+                          shuffle=True,
+                          initial_epoch=1)
+            else:
+                print('[earlystop] Training stopped after 1st epoch!')
+
         else:
             print('Using real-time data augmentation.')
             # This will do preprocessing and realtime data augmentation:
@@ -150,16 +168,35 @@ def main(argumentList):
                 horizontal_flip=True,  # randomly flip images
                 vertical_flip=False)  # randomly flip images
 
-        # Compute quantities required for feature-wise normalization
-        # (std, mean, and principal components if ZCA whitening is applied).
-        datagen.fit(x_train)
+            # Compute quantities required for feature-wise normalization
+            # (std, mean, and principal components if ZCA whitening is applied).
+            datagen.fit(x_train)
+
+            # fit for 1st epoch
+            model.fit_generator(datagen.flow(x_train, y_train,
+                                             batch_size=batch_size),
+                                epochs=1,
+                                validation_data=(x_test, y_test),
+                                workers=4)
 
-        # Fit the model on the batches generated by datagen.flow().
-        model.fit_generator(datagen.flow(x_train, y_train,
-                                         batch_size=batch_size),
-                            epochs=epochs,
-                            validation_data=(x_test, y_test),
-                            workers=4)
+            # Score trained model.
+            scores = model.evaluate(x_test, y_test)
+
+            if scores[1] > acc_break_limit:
+                # If the model looks promising....
+                # Fit the model on the batches generated by datagen.flow().
+                model.fit_generator(datagen.flow(x_train, y_train,
+                                                 batch_size=batch_size),
+                                    epochs=epochs,
+                                    validation_data=(x_test, y_test),
+                                    workers=4,
+                                    initial_epoch=1)
+
+                # Score trained model.
+                scores = model.evaluate(x_test, y_test, verbose=1)
+
+            else:
+                print('[earlystop] Training stopped after 1st epoch!')
 
         # Save model and weights
         if not os.path.isdir(save_dir):
@@ -168,10 +205,8 @@ def main(argumentList):
         model.save(model_path)
         print('Saved trained model at %s ' % model_path)
 
-        # Score trained model.
-        scores = model.evaluate(x_test, y_test, verbose=1)
-        print('Test loss:', scores[0])
-        print('Test accuracy:', scores[1])
+        print('[results] Test accuracy:', scores[1])
+        print('[results] Test loss:', scores[0])
 
         sys.stdout.write(str(scores[1]))
         sys.stdout.flush()