TensorFlow at Google I/O 2018!
By Marcus Chang, Program Manager
Over 7000 people attended I/O this year! TensorFlow was well represented with 7 talks and the AI & Machine Learning sandbox for attendees to explore what’s new!
Sessions featured (you can view the entire playlist here)…
TensorFlow has been extended to simplify model training and deployment using the JavaScript language. Watch this recap to get a detailed description on how to use JavaScript to train and deploy your models.
TensorFlow in production: TF Extended, TF Hub, and TF Serving
In this session we introduced TensorFlow Extended (TFX), TensorFlow Hub, and announced new innovations and features in TensorFlow Serving. As Machine Learning is evolving from experimentation to serve production workloads, so does the need to effectively manage the end-to-end training & production workflow including model management, versioning, and serving. TFX provides this solution to Google and we will outline release plans to deliver TFX to the community. TensorFlow Hub is a central repository of reusable parts of TensorFlow models. With its libraries you can incorporate these parts in your models for transfer learning and package them up to be served with TensorFlow Serving.
Advances in machine learning and TensorFlow
Artificial intelligence affects more than just computer science. Hear a collection of short presentations from top ML researchers. Hear from the TensorFlow engineers working on robotics and the Project Magenta Team who explore the border between machine learning and art.
Get started with TensorFlow’s High-Level APIs
High-level APIs like tf.keras enable developers to train models easily and effectively. This session will introduce these APIs, and notebooks you can run live in the browser to get started using Colab. We’ll walk you through writing your first neural network in TensorFlow using just 10 lines of code with tf.keras, and then we’ll introduce you to Eager execution. We’ll close with educational resources you can use to learn more about ML. By releasing easier and more intuitive APIs, we hope to make TensorFlow, an open-source machine learning framework more accessible for all.
TensorFlow Lite for mobile developers
TensorFlow Lite enables developers to deploy custom Machine Learning models to mobile devices. This technical session describes in detail how to take a trained TensorFlow model, and use it in a mobile app through TensorFlow Lite.
TensorFlow and deep reinforcement learning, without a PhD
On the forefront of deep learning research is a technique called reinforcement learning, which bridges the gap between academic deep learning problems and ways in which learning occurs in nature in weakly supervised environments. This technique is heavily used when researching areas like learning how to walk, chase prey, navigate complex environments, and even play Go. This session teaches a neural network to play the video game Pong from just the pixels on the screen. No rules, no strategy coaching, and no PhD required.
Distributed TensorFlow training
To train Machine Learning models effectively, you need to distribute training jobs to multiple machines in a cluster. TensorFlow offers rich functionality to achieve this. Watch this recap to learn how to set this up.
The AI & Machine Learning sandbox introduced attendees to cool demos built with TensorFlow:
TensorFlow
Our TensorFlow team members were on hand to answer questions and show code! To get started with TensorFlow, go to www.tensorflow.org.
TensorFlow Lite
A lightweight machine learning library and tools for mobile and embedded devices. To get started with TensorFlow Lite, go to: tensorflow.org/mobile/tflite. Code can be found here: github.com/tensorflow/tensorflow.
Magenta
A research project exploring the role of machine learning in the process of creating art and music. It’s an exploration in building smart tools and interfaces that allow creative coders, artists and musicians to extend (not replace!) their processes using these models.
Develop your own interface for creating music with TensorFlow.js and Magenta.js @ magenta.tensorflow.org/js. Explore Latent Spaces for melodies and rhythms with MusicVAE, a machine-learning powered generative tool that enables controllable and expressive variations through Latent Loops.
Donkey Car @ DIYRobocars
An open source DIY self driving platform for small scale cars. If you’re curious about these miniature self-driving cars, you can build your own @ donkeycar.com and race them at diyrobocars.com!
Other machine learning powered demos included:
Storyboard
This Android app uses machine learning to transform your videos into single-page comic layouts! In Storyboard, a machine learning algorithm selects video frames, which are then mapped to panels in a comic layout. Each panel is then cropped, zoomed, and stylized using research on machine perception from Google AI. And it all runs entirely on your device!
Try it out on Google Play!
Semantris
Word association games powered by machine learning. Built with Universal Sentence Encoder, the module encodes text into high dimensional vectors that can be used for text classification, semantic similarity, clustering and other natural language tasks. You can play the games here and learn more about experiments in text understanding here.
Learn with Google AI
Machine Learning Crash Course (MLCC) with TensorFlow APIs, Google’s fast-paced practical introduction to machine learning. MLCC features a series of lessons with video lectures, real-world case studies, and hands-on practice exercises. Free, online, and self-paced! Start your training now @ g.co/mledu/mlcc-io.