On Monday, October 4, local time, Facebook’s parent company Meta announced the launch of a new set of a free software platform for artificial intelligence, supporting both Nvidia as well as AMD chips, facilitating developers to develop artificial intelligence programs to switch back and forth between hardware systems based on different chips.
Meta’s latest release of this AI open source software is built on the PyTorch open source machine learning framework, which allows code to run 12 times faster on Nvidia’s flagship A100 chip and 4 times faster on AMD’s MI250 chip.
Meta said in a blog post that the AI software platform not only improves code running speed but also supports AI chips from different vendors.
Software development has now become a key area for chip makers to build developer ecosystems and use their own chips. For example, CUDA developed by NVIDIA is hugely popular.
However, developers who develop AI code based on Nvidia chips through CUDA have difficulty running it on graphics and image processing chips made by companies such as Nvidia’s rival AMD. Meta says its newly launched AI platform allows developers to easily switch their AI applications between hardware systems based on different chips.
Meta said in a blog post that “unified GPU backend support minimizes application migration costs and helps deep learning application developers choose a wider variety of hardware.”
NVIDIA and AMD did not immediately respond to requests for comment.
The software launched by Meta is said to have been developed specifically for a type of artificial intelligence work known as inference. During inference, developers make quick judgments about images by invoking machine learning algorithms that have been previously trained on large amounts of data.
“This is cross-platform software work that demonstrates the importance of software, particularly the importance of deploying neural networks for inference in machine learning,” said David Kanter, founder of MLCommons, a third-party organization that measures the speed of artificial intelligence processing.
The new AI platform will “give customers more options,” Kanter added.
