Metasaid on Monday that it has launched a new set of free software solutions for artificial intelligence applications that might allow developers to switch between different keychains.
According to Meta, its new open-source AI platform based on PyTorch, an open-source machine learning framework. It can even assist code run up to 12 times faster on Nvidia''s flagship A100 chip or up to four times faster on AMD''s MI250 chip.
In a blog post, Meta described the speed increase as demonstrating the flexibility the software can provide.
The software has become a key source of friction for chipmakers seeking to build up an ecosystem of developers to utilize their chips. Nvidia''s artificial intelligence platform has been the most popular so far.
Even if developers tweak their code for Nvidia chips, it is difficult to run it on graphics processing units or GPUs, from Nvidia competitors, according to Meta. The software is designed to be very simple to swap between chips without being locked in.
"In a blog post, the unified GPU back-end support enables deep learning developers to make more hardware vendor choices, along with minimal migration costs.
- Meta Demands Secret Details From Over 100 Firms in FTC Lawsuit: Details
Nvidia and AMD have not made any calls for information.
Meta''s software is used to make automated inference, a method used to learn advanced algorithms that have previously been trained on large amounts of information to make timely judgments, such as determining whether a photograph is of a cat or a dog.
"This is a multi-platform software effort." David Kanter, the founder of MLCommons, a nonprofit that enables learning to be able to utilise neural networks in machine learning for inference.
Kanter said that this new Meta AI platform would be "good for customer choice."
2022, Thomson Reuters