Google Debuts TensorFlow 2.0 Alpha
At this week’s 2019 TensorFlow Dev Summit, Google announced a major upgrade on the framework, the TensorFlow 2.0 Alpha version. TensorFlow 2.0 focuses on simplicity and ease of use, with updates like eager execution, intuitive higher-level APIs, and flexible model building on any platform.
Last August Google Brain Software Engineer Martin Wicke posted in Google Groups that TensorFlow 2.0 would be a major milestone, which led many in the machine learning community to expect the following upgrades:
- “Eager execution” through alignment of model expectations and practice
- Enhanced compatibility with platforms and languages
- Removal of deprecated APIs to reduce duplication
According to the TensorFlow 2.0 official guide, Google has delivered on the expectations. The new release removes redundant APIs, makes APIs more consistent (Unified RNNs, Unified Optimizers), and better integrates Python runtime with Eager execution.
Here are the major changes:
- API Cleanup: In TensorFlow 2.0, many APIs are either gone or have been moved to other sub-packages. Users can use the v2 upgrade script to apply renames automatically.
- Eager Execution: While TensorFlow 1.X users needed to manually stitch together an abstract syntax tree, TensorFlow 2.0 executes eagerly: “Graphs and sessions should feel like implementation details.”
- No more globals: TensorFlow 2.0 users need no longer rely heavily on implicitly global namespaces.
- A session.run () call is almost like a function call, as shown in the example below:
The overarching goal of TensorFlow 2.0 Alpha is to lower the tech barrier for machine learning, enabling researchers with limited experience or expertise to build models efficiently. For example, easy model building is enabled with Keras, which provides Sequential, Functional, and Subclassing as model-building APIs to enable developers to use the right level of abstraction.
Google is also partnering with online learning platforms such as Fast.ai, Udacity, and Coursera to offer a range of courses, specializations, and degrees for TensorFlow learners. Andrew Ng announced today that Deeplearning.ai ‘s newest specialization course, TensorFlow: From Basics to Mastery, will teach useful TensorFlow practices.
Also announced this week was TensorFlow.js 1.0 for the Javascript community; TF Federated, an open-source framework for machine leaning and other computations on decentralized data; and TF Privacy, a Python-based open source library for fairer and safer training.
There was one more surprise: Google introduced Coral Dev Board, a machine learning platform for building AI products on the edge. The US$149.99 dev board features Edge TPU, a cut-down Google ASIC designed to empower TensorFlow Lite machine learning models and accelerate inference at the edge. Google also released the Coral USB Accelerator for adding the Edge TPU to an existing design, enabling easy integration into any Linux system (including Raspberry Pi boards) over USB 2.0 and 3.0.
TensorFlow is an open source software library for computation developed by the Google Brain team. Its robust machine learning framework has enabled broad usage across many different platforms.
Journalist: Fangyu Cai, Tony Peng| Editor: Michael Sarazen
2018 Fortune Global 500 Public Company AI Adaptivity Reportis out!
Purchase a Kindle-formatted report on Amazon.
Apply for Insight Partner Program to get a complimentary full PDF report.
30s ad
☞ Master Deep Learning with TensorFlow in Python
☞ Intro to TensorFlow - For iOS & Android
☞ Tensorflow and Keras For Neural Networks and Deep Learning
☞ Detect Fraud and Predict the Stock Market with TensorFlow
☞ Artificial Intelligence - TensorFlow Machine Learning
Suggest:
☞ Machine Learning Zero to Hero - Learn Machine Learning from scratch
☞ Introduction to Machine Learning with TensorFlow.js
☞ TensorFlow.js Bringing Machine Learning to the Web and Beyond
☞ Platform for Complete Machine Learning Lifecycle