TensorFlow 2.0 Tutorials

Timeline:

• Jan. 11, 2019: TensorFlow r2.0 preview
• Aug. 14, 2018: TensorFlow 2.0 is coming

TensorFlow 1.0 is out-of-date!

TensorFlow Sucks.

We have TensorFlow 2.0 finally.

Let's get started!

Installation

1. CPU install

pip install tf-nightly-2.0-preview

2. GPU install

pip install tf-nightly-gpu-2.0-preview

Test installation:

In [2]: import tensorflow  as tf

In [3]: tf.__version__
Out[3]: '2.0.0-dev20190129'
In [4]: tf.test.is_gpu_available()
...
totalMemory: 3.95GiB freeMemory: 3.00GiB
...
Out[4]: True

Outline

Lesson1

Simple classification example to show you how to train, load/save checkpoint, evaluate with new TF 2.0.

Lesson 14

4 ways to achieve simple MNIST classifcation with TF 2.0.

1. keras compile & fit
2. forward and backward with keras standard layers.Dense
3. forward and backward with keras subclassing custom layer
4. forward and backward with tf.matmul

Lesson 15

train CIFAR10 dataset with VGG16 network.

Refered Repo.

Our work is not built from scratch. Great appreciatoin to these open works！

• https://github.com/herbiebradley/CycleGAN-Tensorflow
• https://github.com/tensorflow/tensorflow/blob/master/tensorflow/contrib/eager/python/examples/pix2pix/pix2pix_eager.ipynb
• https://github.com/moono/tf-eager-on-GAN