Function Differences with tf.keras.datasets.mnist

tf.keras.datasets.mnist

class tf.keras.datasets.mnist()

For more information, see tf.keras.datasets.mnist.

mindspore.dataset.MnistDataset

class mindspore.dataset.MnistDataset(
    dataset_dir,
    usage=None,
    num_samples=None,
    num_parallel_workers=None,
    shuffle=None,
    sampler=None,
    num_shards=None,
    shard_id=None,
    cache=None
)

For more information, see mindspore.dataset.MnistDataset.

Differences

TensorFlow: The MNIST dataset can be downloaded and loaded using the load_data method inside this class.

MindSpore: Load the MNIST dataset file from the specified path and return the Dataset.

Code Example

# The following implements MnistDataset with MindSpore.
import mindspore.dataset as ds

mnist_dataset_dir = "/path/to/mnist_dataset_directory"
dataset = ds.MnistDataset(dataset_dir=mnist_dataset_dir)

# The following implements mnist with TensorFlow.
import tensorflow as tf

(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()