Thanks for the feedback! I also asked a similar question on the ai stack exchange thread and got some helpful feedback there

It was a great project for brushing up on seq2seq modeling, but I decided to shelve it since someone released a polished website doing the same thing.

The idea was the vocabulary of music composition are chords and the sentences / paragraphs that are measures are sequences of chords or sequences of measures

I think it’s a great project because the limited vocab size and max sequence length are much shorter than what is typical for transformers applied to LLM tasks like digesting novels for example. So for consumer grade harder (12GB VRam) it’s feasible to train a couple different model architectures in tandem

Additionally, nothing sounds bad in music composition, it’s up to the musician to find a creative way to make it sound good. So even if the model is poorly trained, so long as it doesn’t output EOS immediately after BOS, and the sequences are unique enough, it’s pretty hard to find something that isn’t different that still works.

It’s also fairly easy to gather data from a site like iRealPro

The repo is still disorganized, but if you’re curious the main script is scrape.py

https://github.com/Yanall-Boutros/pyRealFakeProducer

Underrated comment

Everyone’s conspiring folks. What’s hard to measure, is who’s conspiring

Meanwhile: NixOS

538s model was a good estimator that year too, they leaned towards Hillary (and to be fair, she did win the popular vote) but certainly kept a trump win in the swing states within margin of error.

270 to win is another good site

Ollama (+ web-ui but ollama serve & && ollama run is all you need) then compare and contrast the various models

I’ve had luck with Mistral for example

I doubt you could put him in prison, he’s still technically a former president, where would you put secret service for example? Lots of undefined legal gray area here

I wanted to be a hacker as a kid, so I had some experience with Backtrack 5. A prof said if you wanted to be a cowboy coder, do everything in your terminal. That was good advice, I’ve learned a lot about OS’s from that

Your OS is basically a set of drivers that allow you to leverage your hardware, as well as a package manager for managing your software, and a system for managing services (like at startup or by some event trigger)

I’m an advanced user but NixOS has been an excellent OS, it’s like all the fun of tuning arch but with less elbow grease. I was a kde neon (ubuntu base + plasma display manager + KDE desktop environment) user before

I’ve tried a few IDEs, mainly Microsoft ones as of recently, but I still prefer my neospacevim setup. Microsoft has a very nice debugger and other useful features for navigating large software projects, but even on my 3080 12th Gen i7 rig with 32GB the plugins I use end up slowing things down. Plus, a similar debugger interface can normally be found in an init.toml layer

With neospacevim, I can specify which plugins get loaded for which file types, so my LaTeX plugins don’t interfere with my Python plugins for example.

Also the macro language locks me into vim, I even installed vimium keybinds for my browser. Spacevim is nice because you can see all the available keybinds option trees by pressing Space.

I mentioned spacevim/SpacEmacs because your post focused on emacs/vim, if you do choose either to make an IDE in I would imagine SpacEmacs/spacevim might be a little closer to an IDE than a text editor.

Spacevim is nice because it will auto install packages declared in the init.toml, sometimes with vanilla vim or neovim you need a plugin manager installed separately

I like Spacevim a lot (inspired by SpacEmacs), you can use neovim as the underlying vim package as well. Then update init.toml with whatever layers/plugins you want