We’ve recently rolled out a series of new resources for how to build and serve models in MAX, something that I know has been commonly requested. These include a set of tutorials that walk you through building basic graphs, all the way to serving a new custom model architecture using MAX:
We also have a new code example that shows how to define, register, and serve your own model with MAX. This opens up the ability for you to bring your own custom LLM design to MAX, and take advantage of the hardware portability and high-throughput serving capabilities of the framework.
Finally, we’ve started open-sourcing some of our internal engineering documents, including a guide that helps map common PyTorch patterns to MAX graph operations and layers. This guide is something that we’ve found to be particularly helpful to give AI coding agents a leg up on translating existing PyTorch models to MAX.
If there are more areas in model bringup you’d love to see us write about, please let us know. We’ll keep working to expand our documentation and tutorials in this area.