78 points by code_monk 6 months ago flag hide 6 comments
username3 6 months ago next
How does the FOSS library compare with alternatives like NumPy and SciPy for core operations? Are there any performance trade-offs for using your library for crunching large numerical datasets?
original_poster 6 months ago next
username3, in comparison to NumPy and SciPy, we aimed for an approachable, user-friendly design that simplifies experimentation and reproducibility for researchers while maintaining reasonable performance. I'd recommend running benchmarks with your specific use case to ensure satisfaction with the performance.
username1 6 months ago prev next
Congrats on building a FOSS machine learning library! That's amazing. I'm curious, what motivated you to create a new library instead of contributing to the existing ones like TensorFlow or PyTorch?
original_poster 6 months ago next
Hi username1, thank you for the kind words and great question. I found the existing libraries to be quite complex, and they didn't cater to the specific needs of my project. Thus, I decided to bootstrap a new library tailored to my project's requirements, and I'm releasing it as open-source software to help others as well.
username2 6 months ago prev next
Impressive work! I'd like to know more about the architecture, especially governance and versioning. What strategies are you employing to ensure the library remains stable while continuing development?
original_poster 6 months ago next
username2, thank you for your support. When it comes to the library's architecture, I focused on keeping the governance lean—similar to Go projects—with fewer but highly skilled contributors. We're following semantic versioning to ensure compatibility and maintain stability while actively developing new features.