 My name is Ibrahim Haddad. I am the exact director of the Pytorch Foundation, and I'm really thrilled to be here today to give you kind of a briefing on the project itself and what we're doing at the Pytorch Foundation since we welcomed the project about a year ago. So the title of the presentation is One Year of Open Governance, since it is almost exactly one year, since September of 2022 when we announced that the Pytorch Project is transitioning to the next foundation from Metta. Before we get into this, I will provide you with a little briefing on how to properly connect a USB. So that worked. A little briefing about the project itself because it is really a fascinating project once you look into it and you see the growth of its community and its deployment and its adoption across various industries. So the project started in 2016 and really what kicked off the initial development on Pytorch was a frustration with previous technologies and previous existing projects that did not allow fast prototyping, they didn't have the existing functionality. So Pytorch was created from that need and it took about two years to become kind of a de facto driving factor in terms of bringing in contributions, thousands of developers, thousands of contributions pouring into the project around 2018. By 2022, the project has gathered so much steam that became the leading project used in research across academia all over the world. And actually I will come back to that because in a later slide because academia is extremely important to Pytorch to the point that as part of the Pytorch Foundation we dedicated a board seat to have academia represented on the board. So that's how strong the connection is between academia and the project. 2022, a year later, Pytorch moved to the Linux Foundation. This is a new era in terms of open governance and I will discuss this a little bit in detail in the coming slides and looking into the future 2024. It's really a lot of opportunities. We have big plans and we're planning to kind of double down on everything we're doing this year and expand going into the future. So a year ago, as I mentioned, same month, actually we were on a podium in Dublin at the Open Source Summit announcing that Pytorch is transitioning to the next foundation. And this was really a monumental announcement just given the importance of the project and especially that the project was already successful and is already successful. The project has hundreds and thousands of contributors, hundreds of thousands of projects that depend on Pytorch as a dependency in GitHub. So it was a significant mark that the project founder and their core partners, the founding members of the Pytorch Foundation as you see them on the screen, they got together and they decided to proceed forward in hosting the project at the foundation in a step to drive additional innovation and additional development in the Open Source ecosystem and putting Pytorch in a new environment will help accelerate that drive and that innovation. So these six organizations as you see on the screen are the founding member of the Pytorch Foundation and they were critical in moving the project into the Linux Foundation. And that transition is important for two different pieces. The first one is basically Metta and the Pytorch founders from their perspective, this is a very important signal to the market or to the industry in general of the importance of Open Source and their commitment to Open Source and Open Governance, even though the project is already under OSI approved Open Source license and there's quite a bit of success around it. And from the Linux Foundation perspective, this is a one step further in our continuous investment in the Open Source AI ecosystem. So for, I see some faces here that are recognized and they're aware of the LF AI and Data Foundation. So we already have an umbrella foundation at the LF today that has over 60 member organization. We host 55 technical projects and we've been investing in supporting that ecosystem since 2018. We see a lot of opportunities or we saw a lot of opportunities in Open Source AI as well as 2017 and beginning 2018 when announced the new umbrella. And today that specific community aggregate since 2018 has over 95,000 contributors into our project and only in 2023 we had 23,000 active contributors in the project. So for us, doubling down on the Python Foundation in that sense and proceeding with the project and supporting it is just another step of what we typically do for the projects and the ecosystems under the Open Source umbrella that we support. So one of the typical question that is your current quite a bit is on the governance side. And I have this slide kind of to clarify that a little bit. We have a complete separation between the project governance which is what we call technical governance and the governance of the foundation itself. So when you think of technical governance this is the technical committee of the project. It consists of the lead maintainers of the project. These individuals are in that position because they are core committers, they are highly skilled in the competence required to drive contributions into the project. They have deep understanding of the project and basically they have a specific set of responsibilities as part of that participation. So this is the technical governance and there's a specific process for any individual who is contributing to the project on how they can build up their portfolio of contributions to drive to become a maintainer or a lead committer as part of the project. And the other aspect of the governance is what you call the foundation governance which is the Pytorch Foundation. And from my perspective I see it as the funding body of the project. So the foundation itself is a mechanism for us to bring in funding that will help us spend on the project whether it's build infrastructure, a website, IT, events, et cetera. So companies come in, join the Pytorch Foundation, they have a yearly membership that they participate in and they contribute to the foundation and they participate in a flurry of activities focused on marketing, communication, events and other activities but there's no interference in how the project run on a daily basis from a technical perspective. And all of the efforts in the foundation are in support of the project and I will talk about this slide here. So as part of the funding we have we gather from the foundation, that funding goes in support of the project. So we have dedicated resources that are hired to support the Pytorch project in the foundation. So there's Kaili here supporting the project from a marketing and communication perspective. I support the project as a GM. We have other resources as well dedicated to supporting the project. So now there are resources that work for the foundation dedicated in support of the project. We are now responsible as a Pytorch Foundation of creating the events of the project. Coming up we have the Pytorch Conference but I know we'll mention that in later slide but other than that we have a lot of different webinars, we have video recordings, we have a lot of educational material in terms of events that were put together in support of the project to bring awareness and to help prepare the next generation of contributors to provide education and bring in people and help them become contributors into the project. We have focused efforts in terms of marketing, PR and communication and you see that a lot from the Pytorch.org. So if you go to the project's website you will see a lot of different activities there in terms of whether it's announcements in relation to training or new releases, meet the maintainers, interviews developers, companies deploying so there's really a lot of use cases being made available. We have integrated the project with our internal Linux Foundation LFX platform which gives us a lot of services in terms of security audits, license compliance audits, project inside understanding the population of developers where they come from which companies which will help us better serve them as we see all these different statistics. And we have really a number of services that are now centralized in the foundation for the project in terms of training, legal support, IT infrastructure is a masterpiece just given how large the project is and the amount of IT and build system it requires. This is actually one of the largest pieces you see on the screen and a number of other services. So previously if you are an open source project without such foundation in support of it all of these different pieces are being provided by different partners, sometimes they're not, sometimes there's kind of disputes on who's gonna take care of what, the cost. Now all of these efforts are being abstracted. These services are available to the project and it's community and it's provided by staff that are dedicated in support of the project. So this is actually, I went last night on Twitter and I'm trying to locate the graph because I have the statistics from one year at the Linux Foundation. So I was able to grab it from Twitter and when PyTorch as a project joined the foundation six months later we posted this in terms of kind of tracking how the project is doing now that's part of the foundation and you will see that across all metrics and kind of the output of existing developers, new developers entering the project, the number of commits, number of people following us on, various social media, number of videos we're able to produce and how many times these videos have been shown, across every single metric we track we've seen an increase and some of these increases are north of 35% which is really incredible that within six months we were able to support the project and help the project grow by that much. By the way all this information now is idealized that actually on the PyTorch website later on. So it was really incredible progress within six months and now when we look again and track all these different metrics for 2023 they're even better. So across all these different metrics and I don't show all of them here just kind of the key ones, you know how many repos on GitHub actually depend on PyTorch as a dependency over 600,000 which is really insane. I think this must be one of the top dependency across all of GitHub. The year over year growth in terms of a new repos being in support of the project. The number of research publications, the number of commits, number of developers. So it's been really an amazing journey for the project and really for us as staff a great journey and very exhaustive just given the amount of activities across all facets of the project. And all of this really together is just an amazing indication that in neutral home for an open source project is really a great way to help the project grow and attract new organizations to participate in the development and be the adopters of that specific technology. And we see that not just in PyTorch but with every single project that you host in the foundation. And in parallel to this and I will give you a little side story here from LFA and data. So in LFA and data which is kind of sister organization to the PyTorch Foundation we host 55 technical projects. We're growing one to two project per month. And one of the things I love to do is when I wanna talk to any given company I always go back to my portfolio of project and see okay, let's say I'm talking to Microsoft or I'm talking to IBM or Intel. I go and see what project these organizations are hosting with us. And I pull in the data from when that project joined us and then see the current set of data in terms of new contributors, number of commits, all these type of statistics. And for some of these projects, they've seen 800% increase in number of developers or number of commits or 900. I've made presentation a couple of weeks ago where most of the projects were in the range of five to 700% increase just within 12 months of joining the foundation which is kind of another data point in support of the hypothesis that when a project transitions to an open source foundation it has an open governance model, its IP and its assets now are managed by the foundation. The foundation staff manages all the aspect of the project from infrastructure to its legal to its marketing and all the other pieces. There's really a lot of confidence being put in that project and organizations looking to adopt and participate in the project have much more confidence in saying, this project is gonna be here in five years regardless whether the founders of the project are here or not. And now this project is a foundation project and I as an organization, I can feel confident in adopting this project and depending on it because I know there's a foundation supporting it and backing it. So that's about the project. In terms of the PyTorch Foundation itself, as I mentioned, we announced the foundation in September of 2022 last year and it took us six to seven months really to establish ourself, create our internal infrastructure, create our processes, our working models internally and then we updated our charter to start welcoming the new organizations into the foundation. So we have similar to many of the umbrella foundations at the LF. We have different levels of membership. We have the premier level of membership for board member, general membership for any other organizations that doesn't wanna join at the board level and then we have a specific level of membership dedicated for educational or academic institution. And this one is actually a fee of charge that are no fees for universities to join us and also they get a board seat which is really incredible. So since we opened up the membership in June, we've had these four organizations join the foundation, IBM, Intel and HikingFace join us at the board level, GraphCore as a general member. And we have a number of other members that we're actually in the process of onboarding into the foundation. And one interesting story about the relation between the project PyTourish and the academia is that once we opened up the membership we've received hundreds, like literally hundreds of applications and when we started inviting them and it's really overwhelming. A lot of these are actually coming from individuals, who are professors at universities or their lab heads at different R&D institutions or even PhD students. That tells you kind of the relation between and the use of PyTourish and academia. So now we're actually in the process of putting together a specific programs that is a membership for individuals. So let's say you're a PhD student, you're a professor and so on and you would like to become a member of the foundation and have that association. So we're kind of tailoring a specific program for them with specific set of benefits that we're aiming to launch next year. So if you're in the audience here, you're interested in learning more about the membership, please grab me after this talk. So back into the project, one of the major milestones of PyTourish was PyTourish 2.0. So for the past year plus there has been a lot of development from PyTourish 1, 1.1 all the way to PyTourish 1.15 or 1.14 and then PyTourish 2.0 kind of blew all the expectations out of the water in a sense where there's almost a guarantee of 50% performance, additional performance or more depending on the specific deployment. So PyTourish 2.0 was released in March 2023. There has been really a lot of coverage for that. Massive improvement in not just new functionalities but increased performance in existing functionalities. And really within a couple of months we've seen most of the major enterprises that we work with have transitioned from 1.x to 2.0. And all the media activities and marketing activities here handled by Kylie has on PyTourish 2.0 kept us busy for a couple of months, really amazing deception and massive amount of improvements in the project. And actually during the Dublin event when we announced the project one of the questions that Jim Zemlin asked me on stage was, was this going to be just PyTourish or is it gonna be like CNCF where you have like Kubernetes but then you have these hundreds of tools and libraries and different enabling technologies around it. And we are actually at that point where we're enabling an ecosystem of tools in support of PyTourish project itself for the goal of enabling and accelerating the use, contribution and adoption of PyTourish. So this year alone we've welcomed so far 18 new tools in support of that PyTourish ecosystem and there are already over a hundred projects that exist today in that library of tools. So if you know, if you're aware of Kubernetes and you're aware of the ecosystem we're actually getting there. We have kind of the core project which is PyTourish and over a hundred different supporting tools and libraries that helps organization bring in, when they adopt PyTourish, bring in a whole set of tools that will have them with their integration and transition. And you know, there are some tools here that were announced this year that you see on the screen but for the full details I would highly recommend going to the URL on the screen which has the listing and you're able to kind of sort and search and do all kind of lookups on the website. So this is actually as thriving the ecosystem as the PyTourish itself in terms of development and activities. So how to get involved? There are really a number of ways you can get involved. As an organization you can join the foundation today. As I mentioned we have a number of levels at the foundation to help work on new organization and if you're in academia or university there's a specific membership class for universities that also allows for a board seat. So I would highly recommend starting there as an organization. As an individual or as someone who works for a company that uses PyTourish on a daily basis one of the things we started doing is creating a PyTourish technical training. So our plan is set to create kind of a number of these where they have a certain curriculum in mind and we started with our first project PyTourish in practice that you see on the screen and we're offering 50% discount on it. And again all the revenues from training things like that actually go in support of the project. So any kind of sales that we make on the training it goes into the PyTourish foundation budget in support of the things like the PyTourish infrastructure and other items. So this is a great way for individuals and we're aiming for a certificate. So basically after achieving a certain number of classes similar to how CNCF does their cloud native certificate so we're following the same model. Another way to get involved is to attend the events. So we have a number of smaller events that we have in different geographies but we also have the anchor major event the PyTourish conference for 2023. And this is one of the areas that the next foundation is transitioning basically a previously meta organized that conference and starting 2023 this event. The PyTourish foundation is the organizer of the conference and we've actually exceeded the potentials of sponsorships so we've welcomed more sponsors than we anticipated. We actually had a cap of 400 registrations we exceeded that and then we had to create and update our contract with the hotel to welcome up to 600 I think. So we opened up additional space for people who would like to attend the conference and it's looking to be an amazing conference so we've had a number of speakers keynoting that are confirmed and if you love Linux Foundation events you're gonna love that event because it's really taking so much effort of everyone to make sure that we have the best PyTourish experience for everyone attending. So that's another way is attending the event socializing with people, figuring out who's doing what and exploring the different opportunities that are made available there. In terms of 2024 we have kind of an idea of where we wanna go basically five core focus areas. The first one is developers. How can we improve the developer journey with PyTourish as a project? How can we provide better tooling for developers to be more efficient? How can we provide and onboard the new developers into the project? How can we improve the number of commits coming from existing developers? So there's this whole category of developer experience, right? Second aspect of focus for 2024 which is a continuation of 2023 is training and certification. We started our first hands-on course for PyTourish. We're developing I think four or five different courses as well leading to a certification. So we haven't announced any of that yet. It's just kind of fork in progress, just FYI for you to know that there's really a big agenda there in terms of providing formal training for developers wanting to enter the PyTourish ecosystem. We have a number of activities that we executed in 2023 that we wanna carry over for 2024 focused on marketing and events. We've had a number of different smaller PyTourish events that were very successful in different geographies. So we're planning to do that more in 2024 and of course continue building up the main conference, the PyTourish conference in 2024. In relation to academia, as I mentioned a little bit earlier, we're planning a specific membership class for academicians so they can join as an individual with a specific set of benefits such as heavy discounts on training, free entrance to the PyTourish events, et cetera. And we are expanding in terms of setting up mentorship programs through our Alifax platform and different areas that are of interest or intersection as well between academia and the project itself. And lastly, we would like to expand enterprise adoption. So we have different programs in support of this goal that we would kick off in 2024. So as you can see, it's like 2024 looking very good in terms of prospect and especially taking the momentum from 2023 and continuing and building on it for 2024. So it's been really an amazing year for us at the PyTourish Foundation. We had the opportunity to work with the who's who in AI in terms of developers. We met the amazing people in terms of organizations and we are here to help and enable and support the developers and their project. Similarly to what every other umbrella foundation at the Alifax is doing. So this is kind of dedicated to PyTourish but here I'm kind of spinning off in a different direction. So if any of you here is focused in their organization on any given open source project that they would consider for hosting any foundation, this would be kind of a great way. This is a great example where you have the top tech companies brought in such an amazing project and even such an amazing project, so over 35% increase in commits and contribution and new developers. So imagine what could that do for other projects. So that's all I had for today. Thank you very much for attending and I'm happy to take any questions you may have. Thank you. Yeah, so Google Cloud is a founding member of the PyTourish Foundation and a lot of people have similar questions. It says, doesn't Google have TensorFlow and why is that? And I think to be very objective on this question, Google Cloud as a business, they wanna do what their customer asked them. So if they have customers and ask them, I want PyTourish support in your cloud, PyTourish cloud as a business division, they wanna go and do that. So there's, I believe there are different camps within Google, a camp that is focused completely on TensorFlow and all its ecosystem and derivative tooling and supporting technologies. And there's also a camp that sees opportunities in other projects as well, even though they're competing or coming from competing organizations and they wanna support that because it drives more business to them. So enabling more revenue streams really is the key here. Okay, did I pass? Okay. What's the hard question then? Or I'm gonna regret that question. Okay, thank you. So we go a long way back, so he's going easy on me. Yes, Mike. We were actually just talking about this, Kylie and myself in the morning. We showed up at the registration desk and I think one thing I would change, like if I just talk about one, it's just resourcing, resourcing. So typically when we spin off a new umbrella, we're not gonna hire too many people, right? We just start with a very small core team and figure out the actual needs of the umbrella foundation and the members and the project and grow from there kind of incrementally. With Spytorch, it was really an avalanche of activities from day one, where we're working with Meta, Google, Microsoft, NVIDIA. These are all large, massive organizations where resources are available and there's an endless supply of money, virtual in the sense that you have a need, you can, you're able to kind of fill it up. In our case, it was like, wow, we're working with all these companies, there's so much work to be done. So I think resourcing is kind of an area where from day one we realized we need more people on deck and we were actively actually looking and hiring and so today we have a dedicated technical, or the technical PM who's dedicated for the project, she's actually a practitioner, she's doing her PhD in AI, so she's been using Spytorch for five, six years now. So she's top natural in the space. We have Kylie focusing on communication, we have a senior marketer focusing on all the marketing programs and focusing on other things and we're able to tap into the links foundation resources for all of these. But I think the other way to answer your question is what were the things that worked really well, right, kind of in a different way? So certainly there was a very strong will from the founder of the project and all the kind of the core contributors that there is a need for us to, if you wanna grow that project and make it kind of the de facto standard in its own category, we need a way to put it in a foundation and so on. So I think from day one, there was this will and intention that we're doing this for that purpose, right? And that made things easy as we went along because everybody's there with a common goal, not competing goals for bringing the project into the foundation. Everybody, all these six companies and of course the links foundation as kind of integral party to that, everybody wanted to make sure that Spytorch not just continues to be successful but becomes even more successful and grows this community and become kind of the dominant machine learning framework or library in the space and we're getting there. I mean, all the statistics are point. Even one exercise I did, but I didn't include it here is if you go to Google trend, right? And you put in PyTorch and let's say you put in TensorFlow just to follow up on Jeff's question, right? You will see that like, TensorFlow is year and PyTorch is up there, right? For the past 24 months or I mean for the past, you can pull up like five years and then just see the past year and you can see the difference. So everybody kind of we're lining up on the same goal that we want this to be successful. And I think where some of the, they say the devil is in details and this is where kind of things got, how to move this, how to transition that but that's kind of logistics. But I think I think the core thing is more. It's not a usual project. So you've got a Linux foundation pump. Once you join the Linux foundation, you have a pump. And this is very typical for a lot of the project. And I think, I mean, the past couple of months I've been talking to a lot of companies hosting projects for us. And one of the things they ask is how our projects are doing. I mean, they know how the projects are doing from their perspective but we can go back to our internal platform and pull up all these data that I showed examples. So a lot of the project are 400, 500% more in terms of, so what that means is for every project you're putting from IBM into the AF360, you're getting like four other resources, right? And some projects are, we have a few projects, maybe like four or five that are 800, 900% increase in commits and contributors which is just fascinating. So there's that continuous. Yeah, absolutely. And I think to the first two points that Jeff mentioned, cause they relate, one of the, I worked for Samsung for six years and I was part of the kind of committee that decides what technology should be adopted for any given product line and my focus was kind of open source. So when you think of companies like that, there's a difference between open source and open governance in a sense where PyTorch has always been open source but it wasn't open governance. The same thing for the other project, like Android may be open source but it's not open governance. And we at Samsung, we had, for every commit we had to do, for every phone we need to ship, we need to go get somebody's approval in Google, it's not open governance, you know? And they can stop our shipment, you know? So open governance is extremely important because it gives a signal to other companies that, hey, this is not controlled by a given company. There's a path for you as a contributor to become on that table that you call the technical ceiling committee or whatever every project calls it differently and that provides massive amount of confidence that me as IBM or Microsoft or any other company, I can have faith in that project and that I can build on top of it and not be concerned what's gonna happen to it two or three years in the future. And I think this is the major value. So right now I'm kind of split but there's no hard percentage. I mean, you need to do what needs to be done so there's no... So I'm kind of dedicated to the AI portfolio at the LF and that compromises LFA and data PyTorch and there are other two couple projects that are hosted at the Linux foundation level that we're gonna pull in eventually and consolidate everything. So that basically leaves us with PyTorch and LFA and data. So most of the staff dedicated to either of these they're exclusive on these meaning, you know, there's Kylie, myself, Jen, et cetera, et cetera. We're either serving this or that because it's the same domain. Some of the people have expertise in AI as practitioners or data scientists and so on. So at this point we're continuing as we are. There are no plans on, I don't know if you're handing into a consolidation between the two. It is... That's what you call it, consolidation. Yeah, this is kind of very true assessment. And I think where I would differ in opinions with you is that at some point we're gonna need a separate person for either of the umbrellas just because of the nature of work and the amount of work actually that is involved. I was having a discussion this morning and I was saying, you know, sometimes I go to the office in the morning and I order lunch and I order dinner because I'm like, you know, I work from an actual office, you know. So I don't have the... I don't go to the kitchen and like from home, from work from home. So I think it goes to the genesis where the founders of the PyTorch Foundation, you know, PyTorch was and is in a different space than all the other portfolio of project that we had in LFA and data. It has different needs, it has different requirements, it has different type of community. Its adoption is very different than everybody else. So the founders of the PyTorch Foundation didn't feel that it's a good fit for PyTorch to be there because it deserves its own foundation given its status and so on. And actually at the beginning I was, I had my doubts about this. I'm like, you know, it's just another project. But, and you know, we host ONIX, we host Aurovoid. These are like 10, 15,000 GitHub stores. There's like explosion of activities. But then once we started getting involved in PyTorch, okay, this is another level. Okay, it's, you know, it's like, you know, LFA and data here in terms of project activities and that thing is over here. There are periods of time where you like, for two days straight you're just focused on PyTorch and everything that's happening there. So I think this split was intentional for a good reason. And ideally it should be all folded under one, but I think the future will decide how these, how these, how these will work. In terms of resourcing, I think eventually there has to be some more focus because LFA and data is growing like crazy. We just added like 10 members in the past couple of months. We've, we're adding almost two projects a month. Our developer community, if you think of a number of projects, if every project, if, you know, have the project have a request every week, then you're done, your time is gone because you need to fulfill the request of the project. You need the different committees you need to manage. You need to prepare the agendas, the decks, that, that, that, and then you're gone. So I think going back kind of to Max's point, resourcing is kind of one of the challenges and Kylie's and everybody. So, yeah, thank you. The topic of SSF around securing the supply chain and there's this massive activity around AI and largely populated with security individuals and new group of AI. And the AI individuals that weren't necessarily without security, you have streams being crossed by the misguides. Yeah, so, so as a background to what Jeff was mentioning, we have OpenSSF and they're looking into security of models and other artifacts in AI. And we in LFA and data, we have a committee called ML Security, MLSecUp, I think, and, you know, they're focused on the reverse aspect of security in AI. So we're setting up a collaboration together and I think we're gonna have an announcement later this week. I'm not sure if it's tomorrow or the day after, but basically we're launching the generative AI commons and focused on generative technologies and LLMs or models in general, under which you have large language models, small and whatever. And part of one of the work stream that the committee decided on what they're gonna focus on is security. So there's gonna be actually more interaction across a different set of companies working on security in AI. Okay, yeah, please. So for feedback on code, it's all on GitHub issues, right? But then for general discussion, there's a discussion forum. Are you aware of it? I don't know the address of it. Okay, so that, you can send me an email and I can follow up with you. I know that there are, are we doing case studies with other successful adoption? I think that that would be a great idea. So there are companies that deploy and they wanna share their experience and they say, you know, this is how we use that in what context and they talk about their, you know, and all of that should be available on the website. And actually, and that's why I started with code because once we launched the foundation in September of last year within days, I had so many questions from different companies in LFA and data using PyTor. Like, hey, you know, we would like to talk with the maintainer if you can arrange kind of, and I'm like, I talked to the maintainers, like, no, just send them to GitHub, you know? So like, you know, if it's good, it's code, just GitHub issues. You can just create an issue, you know, open an issue and then target was whatever topic and then it gets assigned accordingly.