Open Source Program Office
Open Sourcing the Wisconsin Idea
Learning to code can be daunting for some, especially when learning how to start codes in open source. Sebastian Raschka, a former assistant professor in the Department of Statistics at the University of Wisconsin-Madison, is continuously teaching others how to start their very own open-source projects. Currently a staff research engineer at Lightning AI, his work focuses on implementing new large language models and training them. He sits on the advisory board for the University of Wisconsin-Madison’s Open Source Program Office (OSPO).
Raschka was drawn to open source after one of his PhD committee members, Titus Brown, introduced the concept to him. After learning more about open source, Raschka had found his community and has been a contributor ever since.
Raschka has developed a repository illustrating in code how to create a large language model from the ground up. Despite its quantitative success on GitHub with 20,000 stars, the repository can give programmers access to a comprehensible foundation when beginning a project.
“It already had a lot of impact by helping people understand how large language models work by coding it and using the other libraries,” Raschka said. “This is more for educational purposes.”
Documentation with this project is imperative to Raschka. Without it there would be no way to guide those on how to build these large language programs on their own. These can essentially be considered a manual or set of instructions for coders when beginning their open-source project. With a repository that guides users on building their own large language program from scratch, documentation is essential in showing them how to use this code.
Communication with the rest of the contributors and maintainers is equally important to documentation. It is important to understand other contributors’ ideas and values when going into an open-source project, according to Raschka. Once the number of contributors begins to rise in one project, it can become challenging when deciding what direction to take the open-source project.
“Thinking about how it impacts other users basically is also really important. If you change something that changes the usage of the tool, it breaks everyone’s code who’s relying on your software,” Raschka said. “It’s really important to think about making a library better, but in a way that it doesn’t break other people’s code making it backwards compatible. That is also very important.”
In this situation, it is important to remember the long term plans for these open-source projects. Questions that answer what features can make them sustainable for long term use and who your target audience is can help resolve these issues.
“[Open source] helps with building more reliable, better research. Basically by having it reproducible, but also for the next generation to pass on projects in a transparent, reproducible way,” Raschka said. “It also can be a very fruitful environment to collaborate.”
Continuing to create code that is beneficial for the community is always something to be sought after. Programmers who are new to the open source scene may feel intimidated to release their code in fear of it being incorrect or judged by their peers, but it can actually be a great learning opportunity, according to Raschka.
“Not every project you share needs to be super polished and sometimes it can be just the research artifacts that are useful. Who knows, maybe if someone finds a flaw in it, people are worried about that, but I think that’s actually almost a good thing because it helps you to correct your research,” Raschka said. “It also helps you in your work because you learn something new.”
Open source can leave a lasting impact on a community. Repositories like Raschka’s can ensure a new generation of coders can learn from a library that can help build their skills and feel confident when contributing to these communities. An essential part of open source is the community, and helping those around learn how to grow can help grow open source.