Skip to content

Add keras models to top of using_tpu #134

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
MarkDaoust merged 5 commits into from
Oct 22, 2018
Merged

Add keras models to top of using_tpu #134

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
c300780
Add keras models to top of using_tpu
MarkDaoust Oct 22, 2018
a0c49e2
Update site/en/guide/using_tpu.md
lamberta Oct 22, 2018
7589bf3
Update site/en/guide/using_tpu.md
lamberta Oct 22, 2018
d845a42
Update site/en/guide/using_tpu.md
lamberta Oct 22, 2018
f24a936
Update site/en/guide/using_tpu.md
lamberta Oct 22, 2018
File filter

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 16 additions & 7 deletions site/en/guide/using_tpu.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,8 +1,19 @@
# Using TPUs

This document walks through the principal TensorFlow APIs necessary to make
effective use of a [Cloud TPU](https://cloud.google.com/tpu/), and highlights
the differences between regular TensorFlow usage, and usage on a TPU.
Experimental support for Cloud TPUs is currently available for Keras and Colab.
Run Colab notebooks on a TPU by changing the *hardware accelerator* in your notebook settings: *Runtime > Change runtime type > Hardware accelerator > TPU*. The following TPU-enabled Colab notebooks are available to test:

1. [A quick test, just to measure FLOPS](https://colab.research.google.com/notebooks/tpu.ipynb).
2. [A CNN image classifier with `tf.keras`](https://colab.research.google.com/github/tensorflow/tpu/blob/master/tools/colab/fashion_mnist.ipynb).
3. [An LSTM markov chain text generator with `tf.keras`](https://colab.research.google.com/github/tensorflow/tpu/blob/master/tools/colab/shakespeare_with_tpu_and_keras.ipynb)

The above examples are the best way to get started with a cloud TPU.

## TPUEstimator

The remainder of this doc is about using the `TPUEstimator` class to drive a
[Cloud TPU](https://cloud.google.com/tpu/), and highlights
the differences compared to a standard `tf.estimator.Estimator`.

This doc is aimed at users who:

Expand All @@ -15,13 +26,11 @@ This doc is aimed at users who:
* Are interested in porting an existing `Estimator` model to
run on Cloud TPUs

## TPUEstimator

`tf.estimator.Estimator` are TensorFlow's model-level abstraction.
`tf.estimator.Estimator` are a model-level abstraction.
Standard `Estimators` can drive models on CPU and GPUs. You must use
`tf.contrib.tpu.TPUEstimator` to drive a model on TPUs.

Refer to TensorFlow's Getting Started section for an introduction to the basics
Refer to TensorFlow's Getting Started with Estimators section for an introduction to the basics
of using a [pre-made `Estimator`](../guide/premade_estimators.md), and
[custom `Estimator`s](../guide/custom_estimators.md).

Expand Down