Documentation: Updating tutorials to use TensorFlow Datasets (TFDS).

[dynamicwebpaige]

TensorFlow Datasets (TFDS) is a collection of datasets ready-to-use with TensorFlow. Each dataset is defined as a tfds.core.DatasetBuilder, which encapsulates the logic to download the dataset and construct an input pipeline, as well as contains the dataset documentation (version, splits, number of examples, etc.).

The purpose of this issue is to migrate all tutorials to use TensorFlow Datasets, and to be compliant with the TF 2.0 API. Each migrated tutorial must be eager and distribution compatible, with tests, and all associated engineering artifacts.

[piyush01123]

I want to contribute.
Is this referring to imports like

tensorflow/tensorflow/examples/tutorials/mnist/mnist_with_summaries.py

Line 33 in 226398b

from tensorflow.examples.tutorials.mnist import input_data

where we would maybe want to import data from Tensorflow Datasets:

import tensorflow_datasets as tfds
ds_train, ds_test = tfds.load(name="mnist", split=["train", "test"])

Please correct me if I'm wrong.

[dynamicwebpaige]

Hi, @piyush-kgp! Thanks for the offer to help - we appreciate it! 😊

For more information on tf.data.Datasets, check here.

@rsepassi would be the best contact, and is working with @yashk2810 + several other members of the TensorFlow team to get tutorials ready for TF 2.0. Also make sure to follow progress on these two issues for TFDS support and testing: #31, #25367 .

Ryan, Yash - are there any tutorials on GitHub right now that Piyush could help migrate?

[rsepassi]

That’s great, thanks @piyush-kgp. TFDS is now TF 2.0-ready and can be used in the tutorials. @yashk2810 and his colleagues can coordinate work on the tutorials.

[tensorflow-bot]

Are you satisfied with the resolution of your issue?
Yes
No