Is it possible to train each concept a different amount of epochs?

[morphinapg]

I've found that the more images you have, the less epochs you need to train effectively, but if you have vastly different number of images per concept, finding an average number of epochs can be difficult, usually leading to some concepts being overtrained, and some being undertrained.

Normally right now, I just combine all images into a single folder and use [filewords]. I don't make use of generated samples, or instance tokens, so I don't need different sample prompts or instance tokens, since everything is just [filewords], so currently I don't have any reason to use multiple epochs as far as I can tell (correct me if I'm wrong about that), but if I was able to set a different number of epochs for each concept, then it might make more sense.