Finetuning guide and training code

[AmericanPresidentJimmyCarter]

Having a bit of trouble finding this, but I have a lot of ideas about how to finetune this and was looking for instructions on training. I think we can get to songs of any length pretty easily by converting the model to the modified one for outpainting and retraining on 512x1024.

[DenisSergeevitch]

Also want to do a finetune; please share a code if it is possible

[chavinlo]

Hi Jonny
and yes I am also interested

[chavinlo]

I can provide you with some GPU compute if needed
Either distributed (https://github.com/chavinlo/distributed-diffusion one of my side projects)
Or local (A100s) but only for a while

[oxysoft]

Do you guys think we could fine-tune this with dreambooth as usual? It seems unlikely to me that this was trained on copyrighted music, nor will anyone with the compute to do full-blown fine-tuning. But don't worry, if we can get dreambooth working well enough we'll have a bunch of useful idiots like me to take the hit. Would love to do fine-tunes on various game soundtracks and bands.

[hmartiro]

Additional fine tuning and data information has been added to the model-card. This was trained using approaches similar to hugging face examples, but fine-tuning can be achieved with very small datasets using a dreambooth approach. Will look to add tools to make fine tuning on your own data simpler.