MENU

Alif supports ExecuTorch to boost embedded endpoint-AI

Alif supports ExecuTorch to boost embedded endpoint-AI

Technology News |
By Jean-Pierre Joosting



Alif Semiconductor has announced it will support the ExecuTorch framework that was recently introduced by PyTorch.

ExecuTorch enables on-device inference capabilities across mobile and edge devices and will provide users of Alif’s Ensemble® and Balletto™ devices with an even more seamless path to design, optimize, train, and deploy machine learning models directly on endpoint devices. The integration will allow data and decisions to be processed immediately on-device, which enables privacy-aware personalization, reduces power consumption, and significantly increases the flexibility of on-device AI use-cases.

Balletto AI/ML MCUs integrate a Bluetooth Low Energy 5.3 and 802.15.4 wireless subsystem as well as a dedicated network co-processor, enabling connectivity without the need for a separate chip. Employing the same architecture as the Ensemble family, Balletto utilizes the Cortex-M55 CPU plus the Ethos-U55 neural processor for efficient Edge ML processing. The 32-bit MCUS provide 32-bit microcontrollers provide up to 4 MB of integrated memory at a 1:1 NVM/RAM ratio; 46 GigaOPS of edge ML acceleration; 22x Audiomark/MHz compared to Cortex-M4; and a tiny sub-16 mm² package size.

Targeting power-sensitive IoT applications, the Ensemble family of Arm-based 32-bit AI/ML MCUs scale up to: 19 MB integrated memory; 3670 DMIPS combined compute performance; and 250 GigaOPS AI performance.

“Since we launched our Ensemble family of AI-Enabled MCUs into the market three years ago, we have seen a tremendous level of interest in transitioning machine learning closer to the source of the data,” commented Reza Kazerounian, president and co-founder at Alif Semiconductor. “ExecuTorch will significantly broaden the use cases that can be realized by this transition, and Alif is collaborating with PyTorch on bringing this to microcontroller devices.”

Alif is also collaborating with Arm on bringing support for hardware acceleration of transformer-based models for the first time to microcontroller devices, which will enable a frictionless path for language models and other advanced use-cases to be deployed to edge devices using ExecuTorch.

The AI-enabled Ensemble MCUs and fusion processors from Alif Semiconductor are fully mass production qualified

https://alifsemi.com

If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News

Share:

Linked Articles
10s