You probably shouldn't train an LLM in the browser - here's how (vin.how)

I'm curious how this performs on everyone's machines? Worked great for me on macOS but not as well on Linux (supposedly there are still WebGPU quirks?)