Arm is significantly advancing its role in AI and machine learning by forming new partnerships and integrating its AI acceleration tools into key frameworks. The goal of these collaborations is to boost the performance of AI workloads across various platforms, from edge devices to the cloud.

Empowering AI on Arm Hardware

As part of its latest efforts, Arm has integrated its Kleidi AI acceleration technology into popular machine learning platforms, including PyTorch and ExecuTorch. ExecuTorch, the recently developed on-device inference runtime from PyTorch, is specifically designed to optimize AI processing directly on devices. Announced on September 16, 2024, this collaboration is aimed at delivering enhanced AI capabilities by leveraging Arm CPUs, which will enable next-generation applications to handle complex tasks such as large language models (LLMs).

Arm’s broader goal is to bring AI performance advantages not only to edge computing devices but also to cloud environments, making it easier for developers to build and deploy AI solutions that operate efficiently across different platforms. By integrating with PyTorch and TensorFlow, Arm is ensuring that its Kleidi AI libraries are accessible within these widely used frameworks, allowing developers to take advantage of Arm’s computing power without needing to rebuild their applications from the ground up.

Driving AI Performance with Kleidi Technology

The introduction of Kleidi has been a game-changer since its launch, unlocking significant AI performance boosts for Arm CPUs. According to Alex Spinelli, Arm’s vice president of developer technology, Kleidi has already accelerated development and enabled applications to achieve notable improvements in performance. In the cloud, Kleidi enhances existing efforts to optimize PyTorch through the use of Arm Compute Libraries (ACL), which provides a foundation for enhancing AI performance on Arm-based systems.

One of the key benefits of Kleidi is its ability to improve application performance over time, with little additional effort required from developers. As PyTorch releases new framework versions, applications will automatically benefit from performance gains provided by Kleidi’s optimizations, eliminating the need for constant manual updates.

Kleidi’s Three Key Focus Areas

Kleidi’s design revolves around three major principles:

1. Seamless Integration: Arm’s open technology is deeply embedded within major AI frameworks, ensuring that AI models, such as LLMs, can tap into the full performance capabilities of Arm CPUs.
2. Developer Support: Arm provides extensive resources to support developers, including detailed usage instructions, educational content, and hands-on demonstrations to ensure efficient adoption of its technologies.
3. Growing Ecosystem: Arm has cultivated a thriving ecosystem of machine learning providers, open-source projects, and AI platforms. This collaboration helps ensure that developers have access to the latest AI tools and innovations.

Expanding Partnerships and Ecosystem

Arm is actively working with various machine learning stakeholders, from cloud service providers like AWS and Google Cloud to independent software vendors (ISVs) such as Databricks. These collaborations are designed to further enhance AI workloads on Arm CPUs. By building demonstration stacks, Arm is offering clear examples of how developers can deploy AI tasks on Arm hardware, providing a roadmap for optimizing AI workloads at scale.

Additionally, Arm’s KleidiAI library, which focuses on empowering developers working within AI and machine learning frameworks, is set to be integrated into ExecuTorch in October 2024. This upcoming integration is expected to further streamline the process of building and optimizing AI solutions on Arm-based hardware, making the technology even more accessible for developers working in diverse environments.

Arm’s partnership with PyTorch and the integration of Kleidi AI technology represents a significant leap forward in optimizing AI performance on Arm CPUs. By seamlessly integrating into major machine learning frameworks and working closely with cloud service providers and software vendors, Arm is positioning itself as a critical player in the AI ecosystem. These advancements will enable developers to deliver high-performance AI applications from edge devices to cloud environments, making AI more efficient and scalable across the board.