def load_data(self):
cifar10 = datasets.cifar10
(train_images, _), (_, _) = cifar10.load_data()
train_images = train_images.astype('float32')
train_images = data_utils.normalize_inputs(train_images)
train_dataset = tf.data.Dataset.from_tensor_slices(
train_images).shuffle(self.buffer_size).batch(self.batch_size)
return train_dataset
def load_data(self):
fashion_mnist = datasets.fashion_mnist
(train_images, _), (_, _) = fashion_mnist.load_data()
train_images = train_images.reshape(train_images.shape[0], 28, 28,
1).astype('float32')
train_images = data_utils.normalize_inputs(train_images)
train_dataset = tf.data.Dataset.from_tensor_slices(
train_images).shuffle(self.buffer_size).batch(self.batch_size)
return train_dataset
def load_data_with_labels(self):
(train_images,
train_labels), (_, _) = tf.keras.datasets.mnist.load_data()
train_images = train_images.reshape(train_images.shape[0], 28, 28,
1).astype('float32')
train_images = data_utils.normalize_inputs(train_images)
# Batch and shuffle the data
train_dataset = tf.data.Dataset.from_tensor_slices(
(train_images,
train_labels)).shuffle(self.buffer_size).batch(self.batch_size)
return train_dataset