Developers were mentally prepared with the announcement by the Google I/O I may about building a latest version for TensorFlow Lite for mobile phones. And, now the company has released a developer preview for the TensorFlow Lite.
The software library aims to create a lighter machine learning solutions for embedded devices and smartphones. Google calls it as an evolution in the TensorFlow Lite for mobile devices, which is now available for both iOS and Android app developers.
Focus will not be on the training models, but on diverting a low-latency inference to the less robust devices from conventional machine learning models. This means in layman’s terms that TensorFlow Lite will be focusing on applying models’ current capabilities to the new data that is received, rather than grasping new capabilities from traditional data. This is what something that just couldn’t be found in most of the mobile with enough horsepower to handle.
A post released by the team of TensorFlow Lite says that, “With this developer preview, we have intentionally started with a constrained platform to ensure performance on some of the most important common models. We plan to prioritize future functional expansion based on the needs of our users. The goals for our continued development are to simplify the developer experience, and enable model deployment for a range of mobile and embedded devices.”
However, this is not an entire exposer of the version, so there is much more about to be disclosed as modules get added and library takes shape. For now, Google said the TensorFlow Lite is ready and tuned for the natural language processing models and some different vision such as MobileNet, Smart Reply and Inception v3.