Can the training be parallelized in a manner similar to SETI-at-home?
General model training https://github.com/learning-at-home/hivemind
Stable diffusion specific https://github.com/chavinlo/distributed-diffusion
Inference only stable diffusion https://stablehorde.net/