Google Cloud launches OpenXLA, an open source project to optimize ML model development

Google Cloud recently launched the OpenXLA project, a community-based open source ecosystem of ML compilers and framework projects aimed at making ML frameworks easy to use with various hardware backends for faster, more flexible, and more impactful development.

Proprietary software is a brake on innovation in AI and ML, Google advocates an open source approach. Sachin Gupta, Google Vice President and Chief Infrastructure Officer, says in a blog post dedicated to the project:

“At Google, we believe open source software is key to overcoming the challenges associated with inflexible strategies. And as a core contributor to the Cloud Native Computing Foundation (CNCF), we have over two decades of experience working with the community to turn OSS projects into accessible and transparent catalysts for technological progress. We are committed to open ecosystems of all kinds, and this commitment extends to AI/ML – we strongly believe that no company should own AI/ML innovation.”

The OpenXLA project

Developers often run into incompatibilities between frameworks and hardware when building ML solutions.

The OpenXLA project is an ecosystem of open-source, modular and community-driven compilers co-developed by AI/ML leaders including AMD, Arm, Google, Intel, Meta, NVIDIA…It will reduce, optimize and Efficiently deploy ML models from most major frameworks (TensorFlow, PyTorch, and JAX) to any hardware backend, including CPUs, GPUs, and ML ASICs.

The first objectives

The community will begin by collaboratively evolving the XLA Compiler, (Accelerated Linear Algebra), a linear algebra compiler decoupled from TensorFlow, which allows TensorFlow models to be accelerated without necessarily having to modify the source code.

Thus, in the case of the BERT model, it increased performance by 7 times and batch size by 5 times for an MLPerf submission using 8 Volta V100 GPUs.

It will also evolve StableHLO, a set of portable ML computational operations that facilitates the deployment of frameworks on different hardware options, inspired by the MHLO dialect to which it has brought new features, including serialization and versioning.

During the project’s seed phase in 2022, Google engineers will assume responsibility for the technical direction of the project. Collaboration principles, code review processes, and community infrastructure for OpenXLA will be established in 2023 when the project leaves the TensorFlow organization.

Everyone involved in developing or integrating with XLA is welcome to participate in the discussions. To participate, members can request an invitation to join the GitHub organization and SIG Discord, which will be announced later.

For more information :

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *