i also made something like this. it cover 17M entities across tracks albums artists and labels. posted on show hn a few times but it went unnoticed (hate u (joking))
Hey. Needed this for a while, maybe some other hn people into astrophysics will find it useful too. It's a light package for parsing NASA GCN notices into Python data struct. Parsed notices are flat, fully typed and named so it comes easy to work with them with the autocompletion and static analyzer from your favourite IDE.
That's it, not the funniest project for sure but hope that's useful to someone.
Hello! I find these fascinating too. The 128 dimensions I mentioned are learned by the NN, they are latent space and their actual meaning is ultimately unknown. The number itself is somewhat arbitrary as it was selected optimizing memory usage vs data coverage so that the track embeddings could sit on an A100.
Wrote it with the goal of publishing but realized I do not want to go through the hassle of spamming it around to get it out there. Maybe someone else will find it useful? The repo features all the stuff needed for deploying on amazon AWS EC2.