with a focus on usability, developer productivity,

and simple, intuitive APIs.

If you like Keras and you like Eager Execution,

you will love TF 2.0.

And if you're a longtime fan of the original TensorFlow,

you'll still have the low-level control

that you've come to expect.

Over the last few years, we've added a large number

of components to TensorFlow.

And with TF 2.0, these components

are packaged together into a comprehensive platform that

supports machine learning workflows through training

and deployment.

This TensorFlow 2.0 release includes many API changes

such as argument reorders, API symbol renames, and more.

The TF upgrade V2 tool helps the transition

by converting existing TensorFlow 1.12 Python

scripts to TensorFlow 2.0 preview scripts.

Let's dive into collab and see how you would upgrade.

To use the TensorFlow upgrade V2 script, all you need to do

is PIP install the TF nightly preview.

Once that's done, you can preface your command

with an exclamation point.

Here we can see that we're specifying

an in file, text generation dot PUI, and the name of an out

file, so text generation upgrade dot PY.

We hit Shift Enter, and we're immediately

displayed output code showing all

of the conversions that have taken place

due to this upgrade script.

We can take a look at the report dot text,

and we can also check to make sure

that the original script has been modified with compact V1

terms.

That's it.

You successfully upgraded a TensorFlow model.

Some warnings, do not update parts of your code

manually before running the script.

In particular, functions that have had reordered arguments,

like TF dot argmax or TF dot batch to space,

cause the script to incorrectly add keyword arguments,

and they get mismapped.

This script does not reorder arguments.

Instead, it adds keyword arguments

to functions that have had their arguments reordered.

The conversion process is not able to upgrade all functions.

One notable example is TF dot NN dot

com 2d, which no longer takes they use cuDNN on GPU argument.

If the script detects this, it'll

report to standard out and in the report,

and you can fix it manually.

For example, if you have this, you'll

need to change it to this.

Excellent work.

You just learned how to upgrade your legacy TensorFlow

code to TensorFlow 2.0 with the TF 2.0 upgrade script.

If you run into any snags while doing your model conversions,

please let us know by filing an issue on GitHub.

And if you have any feedback on TensorFlow 2.0,

make sure to let the team know by sending an email

to testing@TensorFlow.org.

You can find a link to that in the video description below.

We're excited to hear what you think and happy engineering.