About datasets
#1
by
qpqpqpqpqpqp
- opened
I think you shouldn't care about licenses, you can just create your random dataset of images or videos, it's to train the autoencoder, not even a txt2img model anyway, it is ethical to train on data regardless of their licenses
Edit: Typo
it is ethical to train on data regardless of their licenses
I disagree. Maybe this aligns with your ethics, but it doesn't with mine, and you don't get to dictate my opinion. I will release training code in the future though, probably after I have a successful video version finished. So at that point you can train on whatever you want.
spacepxl
changed pull request status to
closed