Today, I am proud to announce that PyTorch is moving to the Linux Foundation (LF) as a top-level project under the name PyTorch Foundation. The core mission of the Linux Foundation is the collaborative development of open source software. With a governing board of leaders from AMD, Amazon Web Services (AWS), Google Cloud, Meta, Microsoft Azure and NVIDIA, this model aligns with where PyTorch stands today and what it needs to travel forward. The creation of the PyTorch Foundation will ensure business decisions are being made in a transparent and open manner by a diverse group of members for years to come. The technical decisions remain in control of individual maintainers. I’m excited that the Linux Foundation will be our new home as they have notable experience supporting large open-source projects like ours such as Kubernetes and NodeJS. At this pivotal moment, I want to take a look back at how we started, share why we are moving, and what’s ahead.
This January, PyTorch celebrated its 5 year anniversary! I reflected on what it meant to me in this tweet thread, and this conversation with my colleagues Mike Schroepfer, Lin Qiao, and Yann LeCun. When we started PyTorch development in 2016, it was a collective effort by a band of people from the [Lua]Torch community with a big chunk of people and funding from Meta and individuals contributing from NVIDIA, Twitter and other entities.
Since 2017, PyTorch has grown far beyond our initial vision. With over 2,400 contributors who have built nearly 154,000 projects using PyTorch as a foundation, PyTorch has become one of the primary platforms for AI research, as well as commercial production use. We’ve seen its impact across industry and academia, from large companies to numerous university courses at Stanford, NYU, EPFL, Oxford, and other academic institutions. As a maintainer of PyTorch, the journey has been extremely fulfilling, with the impact of the project seen in various fields from self-driving cars to healthcare to aerospace.
As PyTorch grew, many companies have made foundational investments around it. While Meta remains the largest contributor to PyTorch, companies such as AMD, Amazon Web Services (AWS), Google Cloud, HuggingFace, Lightning AI, Microsoft Azure, Nvidia, and many others have made significant investments, including both technical contributions and community building efforts. They’ve established teams around PyTorch or filled significant voids within the PyTorch community and sent countless contributions to the PyTorch core and to the ecosystem around it — PyTorch is an important part of their future. With PyTorch continuing to grow as a multi-stakeholder project, it’s time to move to a broader open-source foundation.
The business governance of PyTorch was fairly unstructured for quite some time since launch – we operated like a scrappy startup. Team members at Meta spent the time and energy to structure this properly and organize PyTorch into an organizationally more healthy entity. Meta helped PyTorch with introducing many structures, such as Contributor License Agreements, Branding Guidelines, and Trademark registration. Keeping PyTorch’s organizational health up to check is essential and beneficial for the community. The next stage of our organizational progress is to support the interests of multiple stakeholders, hence moving to a foundation is good. We chose the Linux Foundation as it has vast organization experience hosting large multi-stakeholder open-source projects with the right balance of organizational structure and finding specific solutions for these projects.
Simultaneously, the technical governance of PyTorch has been a loosely structured community model of open-source development — A set of people maintaining PyTorch by area with their responsibility often tied to their individual identity rather than their employment. While we kept a codified list at the PyTorch – Persons of Interest page, the technical governance was not formalized nor codified. As PyTorch scales as a community, the next step is to structure and codify. The PyTorch Technical Governance now supports a hierarchical maintainer structure and clear outlining of processes around day to day work and escalations. This doesn’t change how we run things, but it does add discipline and openness that at our scale feels essential and timely.
It’s been an exciting journey since 2016. I am grateful for the experiences and people I’ve met along the way. PyTorch started with a small group of contributors which have grown and diversified over the years, all bringing in new ideas and innovations that would not have been possible without our community. We want to continue the open-source spirit – for the community and by the community. Thank you to our contributors, maintainers, users, supporters and new foundation members. We look forward to the next chapter of PyTorch with the PyTorch Foundation.