sampleCode2.py

def do_training(train, train_labels, test, test_labels, num_classes):
    #set TensorFlow logging level to INFO
    tf.logging.set_verbosity(tf.logging.INFO)

    # Build 2 hidden layer DNN with 10, 10 units respectively.
    classifier = tf.estimator.DNNClassifier(
        # Compute feature_columns from dataframe keys using list comprehension
        feature_columns =
            [tf.feature_column.numeric_column(key=key) for key in train.keys()],
        hidden_units=[10, 10],
        n_classes=num_classes)

    # Train the Model
    classifier.train(
        input_fn=lambda:train_input_fn(train, train_labels,100),
        steps=1000
    )

    # Evaluate the model
    eval_result = classifier.evaluate(
        input_fn=lambda:eval_input_fn(test, test_labels,100)
    )

    return (classifier, eval_result)