what feels like many small fixes

talos/autom8/autoparams.py

@@ -1,5 +1,5 @@
 import numpy as np
-from tensorflow.keras.optimizers import Adam, Nadam, Adadelta, SGD
+from tensorflow.keras.optimizers.legacy import Adam, Nadam, Adadelta, SGD
 
 
 loss = {'binary': ['binary_crossentropy', 'logcosh'],

talos/model/normalizers.py

@@ -11,8 +11,8 @@ def lr_normalizer(lr, optimizer):
         The optimizer. For example, Adagrad, Adam, RMSprop.
     """
 
-    from tensorflow.keras.optimizers import SGD, Adam, Adadelta, Adagrad, Adamax, RMSprop
-    from tensorflow.keras.optimizers import Nadam
+    from tensorflow.keras.optimizers.legacy import SGD, Adam, Adadelta, Adagrad, Adamax, RMSprop
+    from tensorflow.keras.optimizers.legacy import Nadam
     from talos.utils.exceptions import TalosModelError
 
     if optimizer == Adadelta:

talos/templates/datasets.py

@@ -20,7 +20,7 @@ def telco_churn(quantile=.5):
 
     df = pd.read_csv('https://raw.githubusercontent.com/autonomio/examples/master/telco_churn/telco_churn_for_sensitivity.csv')
 
-    df = df.drop(['val_acc', 'loss', 'f1score', 'acc', 'round_epochs'], 1)
+    df = df.drop(['val_acc', 'loss', 'f1score', 'acc', 'round_epochs'], axis=1)
 
     for col in df.iloc[:, 2:].columns:
         df = wrangle.col_to_multilabel(df, col)
@@ -34,7 +34,7 @@ def telco_churn(quantile=.5):
         y1 = df.C0.values
         y2 = df.C1.values
 
-    x = df.drop(['C0', 'C1'], 1).values
+    x = df.drop(['C0', 'C1'], axis=1).values
 
     return x, [y1, y2]
 

talos/templates/models.py

@@ -103,12 +103,13 @@ def titanic(x_train, y_train, x_val, y_val, params):
                   metrics=['acc'])
 
     # here we are also using the early_stopper function for a callback
-    out = model.fit(x=x_train,
-                    y=y_train,
+    out = model.fit(x=x_train.astype('float32'),
+                    y=y_train.astype('float32'),
                     batch_size=params['batch_size'],
                     epochs=params['epochs'],
                     verbose=0,
-                    validation_data=(x_val, y_val))
+                    validation_data=(x_val.astype('float32'),
+                                     y_val.astype('float32')))
 
     return out, model
 

talos/templates/params.py

@@ -1,6 +1,6 @@
 def titanic(debug=False):
 
-    from tensorflow.keras.optimizers import Adam, Nadam
+    from tensorflow.keras.optimizers.legacy import Adam, Nadam
 
     # here use a standard 2d dictionary for inputting the param boundaries
     p = {'lr': (0.5, 5, 10),
@@ -34,7 +34,7 @@ def titanic(debug=False):
 
 def iris():
 
-    from tensorflow.keras.optimizers import Adam, Nadam
+    from tensorflow.keras.optimizers.legacy import Adam, Nadam
     from tensorflow.keras.losses import logcosh, categorical_crossentropy
     from tensorflow.keras.activations import relu, elu, softmax
 
@@ -58,7 +58,7 @@ def iris():
 
 def breast_cancer():
 
-    from tensorflow.keras.optimizers import Adam, Nadam, RMSprop
+    from tensorflow.keras.optimizers.legacy import Adam, Nadam, RMSprop
     from tensorflow.keras.losses import logcosh, binary_crossentropy
     from tensorflow.keras.activations import relu, elu, sigmoid
 

talos/templates/pipelines.py

@@ -44,8 +44,9 @@ def titanic(round_limit=2, random_method='uniform_mersenne', debug=False):
 
     '''Performs a Scan with Iris dataset and simple dense net'''
     import talos as ta
-    scan_object = ta.Scan(ta.templates.datasets.titanic()[0],
-                          ta.templates.datasets.titanic()[1],
+
+    scan_object = ta.Scan(ta.templates.datasets.titanic()[0].astype('float32'),
+                          ta.templates.datasets.titanic()[1].astype('float32'),
                           ta.templates.params.titanic(debug),
                           ta.templates.models.titanic,
                           'test',

talos/utils/recover_best_model.py

@@ -53,7 +53,7 @@ def recover_best_model(x_train,
 
         # get the params for the model and train it
         params = df.sort_values(metric, ascending=False)
-        params = params.drop(metric, 1).iloc[i].to_dict()
+        params = params.drop(metric, axis=1).iloc[i].to_dict()
         _history, model = input_model(x_train, y_train, x_val, y_val, params)
 
         # start kfold cross-validation

test-ci.py

@@ -4,17 +4,17 @@
 
     from tests.commands import *
 
+    scan_object = test_scan()
+    '''
     test_latest()
 
-    scan_object = test_scan()
     recover_best_model()
 
     test_random_methods()
 
     test_autom8()
-    test_templates()
-
-    # test_analyze(scan_object)
+    test_templates()''' 
+    test_analyze(scan_object)
 
     test_lr_normalizer()
     test_predict()

tests/commands/test_analyze.py

@@ -20,14 +20,30 @@ def test_analyze(scan_object):
 
     # test the object properties
     r.best_params('val_loss', ['val_acc'])
-    r.correlate('val_loss', ['val_acc'])
+    r.correlate('val_loss', ['val_acc',
+                             'start',
+                             'end',
+                             'activation',
+                             'optimizer',
+                             'losses',
+                             'shapes'])
     r.data
     r.high('val_acc')
     r.low('val_acc')
 
+
+
+
+
     # r.plot_bars('first_neuron', 'val_acc', 'dropout', 'hidden_layers')
     r.plot_box('first_neuron', 'val_acc')
-    r.plot_corr('val_loss', ['val_acc'])
+    r.plot_corr('val_loss', ['val_acc',
+                             'start',
+                             'end',
+                             'activation',
+                             'optimizer',
+                             'losses',
+                             'shapes'])
     r.plot_hist('val_acc')
     r.plot_kde('val_acc')
     r.plot_line('val_acc')