Arm is enhancing its AI workload migration capabilities by combining its AI acceleration expertise with the new on-device inference runtime from PyTorch, enabling seamless conveyance of AI and machine learning workloads to Arm-based hardware.
The newly established association aims to optimize AI performance by bridging the gap between on-premise and cloud-based operations. With the assistance of PyTorch and ExecuteTorch, the next generation of applications is poised to run seamlessly on Arm-based CPUs. Arm has teamed up with PyTorch to seamlessly integrate the Arm Keil libraries directly into their respective frameworks. A library for AI/ML framework builders is slated to be built-in into ExecuTorch by October 2024.
Since its inception four months ago, Kleidi has demonstrated remarkable progress, unlocking key AI efficiency enhancements on Arm CPUs, according to Alex Spinelli, Vice Chairman of Developer Expertise at Arm. Kleidi leverages existing research to integrate the Arm Compute Libraries (ACL) with PyTorch, thereby creating a template for streamlining AI processing on Arm-based architectures. With Kleidi, opportunities arise for profitable enhancements in efficiency whenever novel framework iterations emerge, yet this potential cannot be fully leveraged without proactively building upon the Arm architecture today.
