The Modular Accelerated Xecution (MAX) platform is an integrated suite of AI libraries, tools, and technologies that unifies commonly fragmented AI deployment workflows. MAX accelerates time to market for the latest innovations by giving AI developers a single toolchain that unlocks full programmability, unparalleled performance, and seamless hardware portability.
See here to get started with MAX and when you want to report issues or request features, please create a GitHub issue here.
The Discord community is the best place to share your experiences and chat with the team and other community members.
In the examples directory, you will find code examples for model inference, Jupyter notebooks for an interactive experience learning experience and instructions for how to work with benchmarking and visualization tooling.
MAX is available in both stable and nightly builds. To install either version, follow the guide to install the MAX SDK (also see for the system requirements).
Then clone this repository:
git clone https://github.com/modularml/max.gitIf you installed the nightly build, be sure you switch to the nightly branch,
because the main branch is for stable releases and might not be compatible
with nightly builds:
git checkout nightlyTo show off the full power of MAX, a series of end-to-end pipelines for common AI workloads (and more) are ready to run. As one example, this includes everything needed to self-host the Llama 3 text-generation model. These pipelines are completely written in Mojo, and all code is provided so that they can be customized, built upon, or learned from.
In addition to the end-to-end pipelines, there are many examples that exercise various aspects of MAX, from performing inference using PyTorch and ONNX models to demonstrating command-line tooling capabilities.
You can follow the instructions in the README for each example, notebook or tool you want to run.
Check out the notebooks examples for using MAX Engine 🏎️ for models such as
Q: I get ModuleNotFoundError: No module named 'max' when running an example?
A: Please make sure you run
python3 -m pip install --find-links "$(modular config max.path)/wheels" max-enginein your local python environment. Note that max can also be available in your PATH
For MAX Serving, you can pull our Docker Container from the the public ECR here: https://gallery.ecr.aws/modular/max-serving
public.ecr.aws/modular/max-serving
Thanks for your interest in contributing to this repository! We are not accepting pull requests yet.
However, we welcome your bug reports. If you have a bug, please file an issue here.
If you need support, the Discord community is the best place to share your experiences and chat with the team and other community members.
The Mojo examples and notebooks in this repository are licensed under the Apache License v2.0 with LLVM Exceptions (see the LLVM License).
You are entirely responsible for checking and validating the licenses of third parties (i.e. Huggingface) for related software and libraries that are downloaded.
