How to use this?
mashonoid opened this issue · 4 comments
I would like to try but I am a little lost on how to use this.
Is this to be used with automatic1111's webui? or.... how?
Would really appreciate some help.
Thanks
I would like to try but I am a little lost on how to use this. Is this to be used with automatic1111's webui? or.... how? Would really appreciate some help. Thanks
This code (at least to me) looks like it's still "just" a fork of SD that allows us to use those extra functionalities as defined by the Aesthetic Gradient approach and is a proof of concept via given example embeddings
I guess we'll have to wait for a repo that'll allow us to actually train our own gradients and then we'll see if we can plug them in on AUTOMATIC1111's webui via placing our own scripts in the scripts folder
I've got a fork of AUTOMATIC1111's webui which has the option to use aesthetic gradients embeddings https://github.com/MartinCairnsSQL/stable-diffusion-webui/tree/vicgalle-aesthetic-gradients
Any embeddings you train go in the aesthetic_embeddings folder.
On the settings tab at the bottom change quick settings to "sd_model_checkpoint, aesthetic_embedding, aesthetic_embedding_steps" to be able to select an embedding and step count for training.
I've got a fork of AUTOMATIC1111's webui which has the option to use aesthetic gradients embeddings https://github.com/MartinCairnsSQL/stable-diffusion-webui/tree/vicgalle-aesthetic-gradients Any embeddings you train go in the aesthetic_embeddings folder. On the settings tab at the bottom change quick settings to "sd_model_checkpoint, aesthetic_embedding, aesthetic_embedding_steps" to be able to select an embedding and step count for training.
Thank you for your work and letting us know!
Have you considered making / already made a pull request for the webui with your code?
I've started on a draft pull request AUTOMATIC1111/stable-diffusion-webui#2498
I need to do some more work on integrating it fully with the more advanced prompt generations where the prompt can change at different steps.