For some reason, in almost every test I did with Gemini 3.5 Live, it classified my Swedish (native) as Dutch. There are similarities between the languages, but still.

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

This was neat, well done!

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.

Hey there, thanks! You might already know this, but if you hover your stone around or turn on "Group View" mode you can see solid colored lines between stones. These indicate whether or not they're close enough to prevent cuts, and are really important. Other than that, I think that the uncertainty and need for estimating is kind of intentional - it's a lot more handwavey than normal Go.

Hope that helps!

Luckily I have access to 4x RTX 4090s, so I didn’t have to pay cloud GPU prices directly. If I had, it probably would have added up quite a bit given how many training runs and experiments I ended up doing.

The biggest speed improvement came from changing the note representation when I switched to compound note events: roughly 5× fewer autoregressive passes per note.

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.

Yes, some kind of planning step is on my TODO list. Another thing I want to try is generating a few continuations in parallel, picking the one that looks best, and then continuing from there. Maybe the picking could be automatic.

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

For DPO I only had around 700 preference examples, so not much data at all. That took about 12 minutes to train on a single GPU.

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

If the attribute describes the current note, I would first try adding it as another field/head on the note event. For example:

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

More global things might be better modelled as a control event:

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

The harder part might actually be finding enough good training data with all of those attributes represented consistently.

Thank you.

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.

Cool project!

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

It seems to be something related the moving average calculation. So it is just a glitch on the chart.