Google shares developer preview of TensorFlow Lite | TechCrunch

Developers were flattering psyched by a announcement at Google I/O behind in May that a new chronicle of TensorFlow was being built from a belligerent adult for mobile devices. Today, Google has expelled a developer preview of TensorFlow Lite.

The program library is directed at creating a some-more lightweight appurtenance training resolution for smartphone and embedded devices. The association is job it an enlargement of TensorFlow for mobile and it’s accessible now for both Android and iOS app developers.

The concentration here won’t be on training models though rather on bringing low-latency deduction from appurtenance training models to reduction strong devices. In layman’s terms this means TensorFlow Lite will concentration on requesting existent capabilities of models to new information it’s given rather than training new capabilities from existent data, something many mobile inclination simply don’t have a horsepower to handle.

Google minute that a large priorities when they designed TF Lite from blemish was to stress a lightweight product that could initialize fast and urge indication bucket times on a accumulation of mobile devices. TensorFlow Lite supports a Android Neural Networks API.

This isn’t a full release, so there’s still many some-more to come as a library takes figure and things get added. Right now Google says TensorFlow Lite is tuned and prepared for a few opposite prophesy and healthy denunciation estimate models like MobileNet, Inception v3 and Smart Reply.

“With this developer preview, we have intentionally started with a compelled height to safeguard opening on some of a many critical common models,” a post authored by a TensorFlow group read. “We devise to prioritize destiny organic enlargement formed on a needs of a users. The goals for a continued growth are to facilitate a developer experience, and capacitate indication deployment for a operation of mobile and embedded devices.”

Interested developers can puncture into the TF Lite documentation and get to obsessing.

More tabs ...

Posted in
Tagged . Bookmark the permalink.
short link