The base model already have a good grasp of language, so that man pages look like coherent sentences. It also knows python and similar language well, which means that things like 1+5 or print(["hi" for _ in range(5)]) work.

I did try to train from scratch, but it was much weaker (also added a lot of aux losses).

``` BOS SCREEN_START 32 96 ```

I trained it on three different resolutions, but it should kind of work with in between ones too.

Gemini 3 Flash was around 0.403, 2.5 Pro was around 0.34.

It’s pretty good at understanding my broken Mandarin, though.

The only alternative version of Go I’ve played before was magnetic Go, where all the rules are the same except that, along the axis lines where you place a piece, like colours attract and opposite colours repel.

On Voronoi, I did feel like the continuous space made it really hard to judge whether there was enough room, whether the opponent could get in, etc. Maybe it would work better if it were discretized a bit? Or maybe I just need to play it more.

For the current model I’m using Core ML, which optimizes the kernels the first time you run it. I haven’t actually spent that much time tuning performance beyond that.

I can probably squeeze out quite a bit more than 100 notes/sec as well. I haven’t spent much time optimizing inference yet.

Pretraining was obviously a a lot slower, the 125M model took roughly half a day.

  NOTE(
    pitch,
    delta_onset,
    duration,
    attack_velocity,
    release_velocity,
    ...
  )

  NOTE(...)
  | CONTROL_SUSTAIN(delta_onset, value)
  | CONTROL_TEMPO(delta_onset, bpm)
  | CONTROL_PROGRAM(delta_onset, instrument)

Yes, I think I’ve gotten it to roughly a GPT-2 level: good enough to share, but with a lot of room left to improve. I think adding some kind of bar/measure token might help with rhythm, and perhaps some form of longer-term planning for the overall composition.

The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device.

The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.

Comments URL: https://news.ycombinator.com/item?id=49373456

Points: 598

# Comments: 117

Comments URL: https://news.ycombinator.com/item?id=48411484

Points: 4

# Comments: 1

That said, for `createAliasMap`, don't you think you could create a deterministic mapping from and to UUIDs <-> word chains? That way, no additional state would be needed. [Might require fairly long word chains...]

I noticed that if you go from training to watch and then back, the training temporarily drop significantly in score.