added an option to save the best model among the all epochs

autonomio/_utils/callbacks.py

@@ -1,6 +1,7 @@
 from keras.callbacks import EarlyStopping
 from keras.callbacks import ReduceLROnPlateau
 from keras.callbacks import LearningRateScheduler
+from keras.callbacks import ModelCheckpoint
 
 from autonomio._utils.get_method import get_method
 
@@ -45,4 +46,10 @@ def step_decay(epoch):
 
         l.append(LearningRateScheduler(step_decay))
 
+    if para['save_best'] is True:
+
+        path = './' + para['save_model'] + '.h5'
+        l.append(ModelCheckpoint(path, save_best_only=True,
+                                 save_weights_only=True))
+
     return l

autonomio/commands/_wrapper.py

@@ -17,6 +17,7 @@ def train(X=None, Y=None, data=None,
           activation_out='sigmoid',
           metrics=['accuracy'],
           save_model=False,
+          save_best=False,
           neuron_max='auto',
           batch_size=10,
           verbose=0,
@@ -134,13 +135,16 @@ def train(X=None, Y=None, data=None,
               https://keras.io/metrics/
 
     save_model =  An option to save the model configuration, weights
-                  and parameters.
+                  and parameters from last epoch.
 
                   OPTIONS:  default is 'False', if 'True' model
                             will be saved with default name ('model')
                             and if string, then the model name
                             will be the string value e.g. 'titanic'.
 
+    save_best = When True saves the best model. Works only when 'save_model'
+                is activated.
+
     neuron_max = The maximum number of neurons on any layer.
 
     neuron_last = How many neurons there are in the last layer.
@@ -286,7 +290,8 @@ def train(X=None, Y=None, data=None,
                   'lr_scheduler': lr_scheduler,
                   'initial_lr': initial_lr,
                   'drop': drop,
-                  'drop_each': drop_each
+                  'drop_each': drop_each,
+                  'save_best': save_best
                   }
 
     if model is 'lstm':

test_script.py

@@ -97,7 +97,7 @@
 tr = train(1, 'quality_score', temp, flatten=.5, learning_rate=0.1)
 tr = train(1, 'quality_score', temp, flatten='mean', metrics='accuracy')
 
-tr = train('text', 'neg', temp, save_model='test_model')
+tr = train('text', 'neg', temp, save_model='test_model', save_best=True)
 te = predictor(temp, 'test_model')
 
 tr = train(1, 'neg', temp, layers=1, validation=True)