Arm’s Latest Partnerships Drive AI Performance from Edge to Cloud
Arm is making significant strides in bringing AI and machine learning workloads to Arm-based hardware by integrating its Arm Kleidi AI acceleration technology with PyTorch and ExecuTorch, an on-device inference runtime developed by PyTorch. This collaboration, announced on September 16, aims to extend AI performance improvements across both edge and cloud environments. By supporting PyTorch and ExecuTorch, the next generation of applications will be able to run large language models (LLMs) on Arm CPUs, unlocking new possibilities for AI-driven solutions. Arm’s partnerships with PyTorch and TensorFlow also ensure that Kleidi libraries are directly integrated into these leading AI frameworks.
KleidiAI, a library designed for AI and ML framework developers, is expected to be integrated into ExecuTorch by October 2024. Since the launch of Kleidi just four months ago, Arm has seen substantial gains in AI performance on its CPUs. According to Alex Spinelli, vice president of developer technology at Arm, Kleidi has already accelerated AI development and provided significant performance uplifts. In the cloud, Kleidi builds on the existing optimization work with the Arm Compute Libraries (ACL), offering a clear blueprint for enhancing AI on Arm platforms. The integration of Kleidi allows applications to benefit from performance improvements with minimal additional effort, as new framework versions are released.
Kleidi is built around three foundational pillars designed to improve AI performance and developer experience. First, it offers open Arm technology that is seamlessly integrated into key AI frameworks, enabling LLMs to harness the full potential of Arm CPUs. Second, Kleidi provides developers with a robust set of resources, including detailed usage guidance, learning paths, and demonstrations, making it easier to adopt and leverage the technology. Finally, Kleidi is supported by a growing ecosystem of machine learning software providers, frameworks, and open-source projects, ensuring that developers have access to the latest AI features and innovations.
These strategic partnerships between Arm, PyTorch, and TensorFlow signal a new era of AI development, where performance is not constrained by the limitations of existing hardware. By integrating advanced AI capabilities into Arm’s architecture, these collaborations are set to empower developers to push the boundaries of AI applications, from edge devices to cloud-based solutions.