Apple moved to its own silicon computer chips three years ago, taking a bold step towards total control of its technology stack. Today, Apple has launched MLX, an open source framework specifically designed to run machine learning on Apple’s M-series CPUs.
Most AI software development currently takes place on open source Linux or Microsoft systems, and Apple does not want its thriving developer ecosystem to be left out of the latest big thing.
MLX aims to solve the long-standing compatibility and performance issues associated with Apple’s unique architecture and software, but it’s more than just a technical game. MLX offers a user-friendly design, likely inspired by popular frameworks such as PyTorch, Jax and ArrayFire. Its introduction promises a more streamlined process for training and deploying AI learning models on Apple devices.
Architecturally, MLX is distinguished by its unified memory model, where arrays exist in shared memory, allowing operations across supported device types without duplication of data. This feature is critical for developers seeking flexibility in their AI projects.
In short, unified memory means that your GPU shares its VRAM with the computer’s RAM, so instead of buying a powerful PC and then adding a beefy GPU with lots of vRAM, you can just use your Mac’s RAM for everything.
However, the road to AI development on Apple silicon has not been without its challenges, mainly due to the closed ecosystem and lack of compatibility with many open source development projects and their widely used infrastructure.
“It’s exciting to see more tools like this for working with tensor-like objects, but I really wish Apple would make it easier to port custom models in a high performance way,” said one developer on Hacker News discussing the announcement.
Until now, developers have had to convert their models to CoreML to make them run on Apple. This reliance on a translator is not ideal. CoreML focuses on converting existing machine learning models and optimising them for Apple devices. MLX, on the other hand, is about creating and running machine learning models directly and efficiently on Apple’s own hardware, providing tools for innovation and development within the Apple ecosystem.
MLX has performed well in benchmark tests. Its compatibility with tools such as Stable Diffusion and OpenAI’s Whisper is a significant step forward. In particular, performance comparisons show the efficiency of MLX, which outperforms PyTorch in terms of image generation speed at higher batch sizes.
For example, Apple reports that it takes “about 90 seconds to fully generate 16 images with MLX and 50 diffusion steps with classifier-free guidance, and about 120 seconds for PyTorch”.
As AI continues to evolve at a rapid pace, MLX represents a critical milestone for Apple’s ecosystem. It not only addresses technical challenges, but also opens up new opportunities for AI and machine learning research and development on Apple devices – a strategic move given Apple’s divorce from Nvidia and its own robust AI ecosystem.
MLX aims to make Apple’s platform a more attractive and viable option for AI researchers and developers, and means a merrier Christmas for AI-obsessed Apple fans.
