why? It's not an obligation. People will put models used and mixed if they want. We should enjoy what we get for free, :).

@davizca some models have restrictive licenses attached to them btw i.e seekart and dreamlike

@syx0 Ah yeah, that's the big joke too. Most "copyrighted/restricted" models are NAI based and most of all (if not all) were trained on MJ images. It's nonsense... We want AI to be open source but we keep putting ridiculous limits. We're not better than Karla Ortiz (at least the one doing this). In the future don't expect people telling on which things trained their model, at least I'm not gonna say a word, being mine works or any others.

Well, since it's 7+gb it's certainly using the unpruned "full" NAI (I.e, uncompressed and filled with loads of latent space hogging up memory for no reason) as a base. Downloading mergers are pointless anyways.

To the author, while I haven't downloaded this a tip is not use the unpruned model for mergers since it's identical to the pruned one for useage, the only difference is that it was still being trained thus not pruned. You can get identical results with the 4gb and the 8gb models. The remaining space is just empty. The only reason models are unpruned in training is to keep them lossless, I.e to not overwrite something when resuming training. Seeing any mix that's larger than 4gb is always a warning flag.