Overview
- Google has announced that it’s working on TensorFlow 2.0 and plans to release it later this year
- The major focus for this version is on usability
- A conversion tool will come with the release to help you update the codes in Python
- This article highlights some of the important features you can expect in this new version
Introduction
TensorFlow is one of the most popular and widely used machine learning frameworks in the industry. While it has it’s limitations and PyTorch has definitely given it a run for it’s money, TensorFlow has embedded itself into many a product, and continues to do so unabated.
Developed and pioneered by Google, TensorFlow v1.0 was launched last year and since then we have seen many releases of this library. Each version has been incrementally better than the previous one, with a number of improvements and bug fixes.
v2.0 has been eagerly awaited for quite a while now and we finally have some good news for you. TensorFlow 2.0 is officially in the works, and Google has released the first details around it this week.
As mentioned by one of the Google Brain Engineers, Martin Wicke, here is what we can expect from TensorFlow 2.0:
- A central feature for this new version will be Eager execution. This is expected to make TensorFlow easier to learn and apply
- Support for more platforms and languages, and improved compatibility and parity between these components via standardization on exchange formats and alignment of APIs
- Removal of deprecated APIs to reduce duplication, which has been the cause of confusion for users
Wicke also mentioned that TensorFlow 2.0 needs to go through a public review process and hence a series of public design reviews (that will cover the planned changes) will be conducted. After the process, the community can give feedback and propose changes for the version.
In order to simplify the transition for current users, a conversion tool will be created which updates the codes in python to use TensorFlow compatible APIs. Otherwise, it will issue a warning in case the conversion is not automatically possible.
Another major change as a part of TensorFlow 2.0 will be that the distribution to tf.contrib will be ceased. For each currently existing contrib module, some will be integrated into the core project or moved to a separate repository while others will simply be removed. For any queries you can get in touch with the team directly by at [email protected]. You can also subscribe to [email protected] for regular updates.
Our take on this
Exciting news! Like most of the ML community, I have been waiting for this version for a while now. While the developer team hasn’t exactly announced too many features yet, it’s still something to go on.
I personally like the conversion tool that will be bundled with the release. It proved to be extremely helpful at the time of the the version 1.0 release as well.
Subscribe to AVBytes here to get regular data science, machine learning and AI updates in your inbox!
An avid reader and blogger who loves exploring the endless world of data science and artificial intelligence. Fascinated by the limitless applications of ML and AI; eager to learn and discover the depths of data science.
