Parallelize AI code with Metal: This is how Apple API works

The sample project

Apple’s Core ML framework uses the GPU via the Metal API and thus enables apps with artificial intelligence on iPhones, iPads and Macs without additional computing capacities in the cloud. The Xcode project for this article, which you can download from the web code at the end, demonstrates this using the example of a Large Language Model (LLM; see glossary on page 134). The open source project contains an LLM in Swift that executes the CPU sequentially, as well as two other versions in Metal for the GPU.

That was the reading sample of our Heise Plus article “parallel to Metal: This is how Apple API”. With a Heise Plus subscription you can read the whole article.