News

TensorFlow Lite will get machine studying mannequin maker

tensorflow-lite-will-get-machine-studying-mannequin-maker

Google has introduced the TensorFlow Lite Mannequin Maker. TensorFlow Lite is an open-source deep studying framework for on-device inference. The brand new software is designed to adapt machine studying fashions to datasets with switch studying. 

“It wraps the complicated machine studying ideas with an intuitive API, so that everybody can get began with none machine studying experience. You’ll be able to practice a state-of-the-art picture classification with solely four strains of code,” Khanh LeViet, developer advocate at Google, wrote in a weblog submit

The Mannequin Maker helps fashions accessible on the TensorFlow hub such because the EfficientNet-Lite fashions. As well as, it helps picture classification and textual content classification. The group plans to supply extra assist for laptop imaginative and prescient and pure language processing use instances. 

Google has additionally added new fields within the metadata to simplify on-device machine studying. The metadata fields fall underneath: machine-readable parameters and human-readable parameters. 

Different updates the corporate has made to the open-source challenge embrace new pretrained fashions, a code generator to generate wrapper code, benchmark instruments to measure mannequin efficiency of fashions, and assist for extra platforms. 

Going ahead, the group plans to launch up-to-date on-device fashions, publish new tutorials and examples, improve Mannequin Maker to assist extra duties, broaden metadata and codegen instruments, and launch extra platform integration. 

“TensorFlow Lite is the official framework to run inference with TensorFlow fashions on edge gadgets. TensorFlow Lite is deployed on greater than four billions edge gadgets worldwide, supporting Android, iOS, Linux-based IoT gadgets and microcontrollers,” LeViet wrote. 

0 Comments

admin

    Reply your comment

    Your email address will not be published. Required fields are marked*