90f779c858aadaeb2b0621ba23e57967ef27d4d2,tests/estimators/classification/test_common_deeplearning.py,,test_fit_image_generator,#Any#Any#Any#Any#Any#,49

Before Change

    classifier, sess = get_image_classifier_list(one_classifier=True, from_logits=True)

    if classifier is not None:
        data_gen = image_data_generator(sess=sess)

        (_, _), (x_test_mnist, y_test_mnist) = get_default_mnist_subset

        true_class = np.argmax(y_test_mnist, axis=1)

        predictions = classifier.predict(x_test_mnist)
        prediction_class = np.argmax(predictions, axis=1)
        pre_fit_accuracy = np.sum(prediction_class == true_class) / x_test_mnist.shape[0]

        np.testing.assert_array_almost_equal(
            pre_fit_accuracy, 0.32, decimal=0.06,
        )

        classifier.fit_generator(generator=data_gen, nb_epochs=2)
        predictions = classifier.predict(x_test_mnist)
        prediction_class = np.argmax(predictions, axis=1)
        post_fit_accuracy = np.sum(prediction_class == true_class) / x_test_mnist.shape[0]

        np.testing.assert_array_almost_equal(
            post_fit_accuracy, 0.68, decimal=0.06,
        )


def test_loss_gradient(
        framework,
        is_tf_version_2,
        get_default_mnist_subset,

After Change

def test_fit_image_generator(
        framework, is_tf_version_2, get_image_classifier_list, image_data_generator, get_default_mnist_subset
):
    try:
        if framework == "tensorflow" and is_tf_version_2:
            return

        classifier, sess = get_image_classifier_list(one_classifier=True, from_logits=True)

        data_gen = image_data_generator(sess=sess)

        (_, _), (x_test_mnist, y_test_mnist) = get_default_mnist_subset

        true_class = np.argmax(y_test_mnist, axis=1)

        predictions = classifier.predict(x_test_mnist)
        prediction_class = np.argmax(predictions, axis=1)
        pre_fit_accuracy = np.sum(prediction_class == true_class) / x_test_mnist.shape[0]

        np.testing.assert_array_almost_equal(
            pre_fit_accuracy, 0.32, decimal=0.06,
        )

        classifier.fit_generator(generator=data_gen, nb_epochs=2)
        predictions = classifier.predict(x_test_mnist)
        prediction_class = np.argmax(predictions, axis=1)
        post_fit_accuracy = np.sum(prediction_class == true_class) / x_test_mnist.shape[0]

        np.testing.assert_array_almost_equal(
            post_fit_accuracy, 0.68, decimal=0.06,
        )
    except NotImplementedError as e:
        warnings.warn(UserWarning(e))


def test_loss_gradient(
        framework,
        is_tf_version_2,
        get_default_mnist_subset,

Instances