added ability to limit permutations by an integer value

.travis.yml

@@ -13,7 +13,7 @@ before_script:
  - sh -e /etc/init.d/xvfb start
  - sleep 3
 script:
- - coverage run --source=hyperio ./test_script.py
+ - coverage run --source=talos ./test_script.py
 after_success: coveralls
 
 deploy:

talos/scan.py

@@ -1,4 +1,5 @@
 from keras import backend as K
+from tensorflow import get_default_graph, Session
 
 from .utils.validation_split import validation_split
 
@@ -10,7 +11,6 @@
 from .utils.estimators import time_estimator
 from .parameters.handling import param_format, param_space, param_index, round_params
 from .parameters.permutations import param_grid
-from .utils.save_load import save_model
 from .metrics.score_model import get_score
 from .utils.pred_class import classify
 from .utils.last_neuron import last_neuron
@@ -23,10 +23,9 @@ class Scan:
 
     def __init__(self, x, y, params, dataset_name, experiment_no, model,
                  val_split=.3, shuffle=True, search_method='random',
-                 save_best_model=False,
                  reduction_method=None, reduction_interval=100,
                  reduction_window=None, grid_downsample=None,
-                 reduction_metric='val_acc',
+                 reduction_metric='val_acc', round_limit=None,
                  talos_log_name='talos.log', debug=False):
 
         self.dataset_name = dataset_name
@@ -50,7 +49,6 @@ def __init__(self, x, y, params, dataset_name, experiment_no, model,
         self.grid_downsample = grid_downsample
         self.val_split = val_split
         self.shuffle = shuffle
-        self.save_model = save_best_model
 
         self.p = param_format(self)
         self.combinations = param_space(self)
@@ -61,6 +59,8 @@ def __init__(self, x, y, params, dataset_name, experiment_no, model,
         self.round_counter = 0
         self.peak_epochs = []
         self.epoch_entropy = []
+        self.round_limit = round_limit
+        self.round_models = []
 
         self.x = x
         self.y = y
@@ -73,8 +73,13 @@ def __init__(self, x, y, params, dataset_name, experiment_no, model,
         self = prediction_type(self)
 
         self.result = []
-        while len(self.param_log) != 0:
-            self._null = self._run()
+
+        if self.round_limit != None:
+            for i in range(self.round_limit):
+                self._null = self._run()
+        else:
+            while len(self.param_log) != 0:
+                self._null = self._run()
 
         self = result_todf(self)
         self.peak_epochs_df = peak_epochs_todf(self)
@@ -104,13 +109,9 @@ def _run(self):
             if self.reduction_method == 'spear':
                 self = spear_reducer(self)
 
-        if self.save_model == True:
-            save_model(self)
-
         K.clear_session()
         self.round_counter += 1
 
-
     def _model(self):
 
         return self.model(self.x_train,

test_script.py

@@ -31,7 +31,8 @@ def iris_model(x_train, y_train, x_val, y_val, params):
                     activation=params['last_activation']))
 
     # here are using a learning rate boundary
-    model.compile(optimizer=params['optimizer'](lr=lr_normalizer(params['lr'], params['optimizer'])),
+    model.compile(optimizer=params['optimizer'](lr=lr_normalizer(params['lr'],
+                  params['optimizer'])),
                   loss=params['losses'],
                   metrics=['acc'])
 
@@ -94,37 +95,37 @@ def cervix_model(x_train, y_train, x_val, y_val, params):
 x, y = ta.datasets.iris()
 
 h = ta.Scan(x, y,
-               params=p,
-               dataset_name='testing',
-               experiment_no='000',
-               model=iris_model,
-               grid_downsample=.0001,
-               reduction_method='spear',
-               reduction_interval=5)
+            params=p,
+            dataset_name='testing',
+            experiment_no='000',
+            model=iris_model,
+            grid_downsample=.0001,
+            reduction_method='spear',
+            reduction_interval=5)
 
 r = ta.Reporting('testing_000.csv')
 
 # here use a standard 2d dictionary for inputting the param boundaries
 
 x, y = ta.datasets.cervical_cancer()
 p = {'lr': (0.5, 5, 10),
-     'first_neuron':[4, 8, 16, 32, 64],
-     'hidden_layers':[2,3,4,5],
+     'first_neuron': [4, 8, 16, 32, 64],
+     'hidden_layers': [2,3,4,5],
      'batch_size': (2, 30, 10),
      'epochs': [3],
      'dropout': (0, 0.5, 5),
-     'weight_regulizer':[None],
+     'weight_regulizer': [None],
      'emb_output_dims': [None],
      'optimizer': [Nadam],
      'loss': [logcosh, binary_crossentropy],
-     'activation':[relu],
+     'activation': [relu],
      'last_activation': [sigmoid]}
 
-ta.Scan(x, y, save_best_model=False,
-           grid_downsample=0.001,
-           params=p,
-           dataset_name='cervix',
-           experiment_no='a',
-           model=cervix_model, reduction_method='spear', reduction_interval=5)
+ta.Scan(x, y,
+        grid_downsample=0.001,
+        params=p,
+        dataset_name='cervix',
+        experiment_no='a',
+        model=cervix_model, reduction_method='spear', reduction_interval=5)
 
 ta.Reporting('cervix_a.csv')