Please tell more about the model
Is this model a mix of models, dreambooths, continued training of particular model?
What model it's based upon, SD v1.4 / v1.5, other?
Can you please specify the content? What's the training set?
Examples of generated pictures?
Strengths and weakness of the model?
Do you recommend to mix it?
would also like to know these things
I really like general model. And like to know more as well.
Here is a generated picture made with this model. Sorry, it a bit horny.
We really need to create some standard test to evaluate any new model with the same set of prompts/negatives/settings, on a variety of styles.
It matches the hash of a mix I already had, so unless I understand hashes wrong it's not continued training.
Okay, I clearly don't understand hashes. I just ran the same prompt through my old mix and Elysium, matching hashes, different pictures.
someone said it's got the same hash as f222?
someone said it's got the same hash as f222?
Yes, it doesn't mean anything as there is a "bug" or wiredness when you merge models with specific conditions you will get the same hash. I run test just to make sure, and f222 and this produce WILDLY different results. So at least we know that the model is based on F222, to no surprise, as the model is very adept at diffusing female form.
weiui get the same hash is not a bug, is to quickly calculate the hash so only a small part of the file header is used.
It is possible that the header of file is same when appending training to an existing model or using add different to mix model.
So it will make same hash, If you calculate the hash of the complete file, you can see the difference between the two files.
You can see this link below for details.
https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/271