About Textual Inversion Embedding

[YouPassTheButter]

Hi, great job! Would you mind sharing the code or pipeline for training the textual inversion embedding?

[shunk031]

Hi, @VSAnimator

I'd like to know about training textual inversion too. In particular, I'd appreciate details regarding the FineTuneConcept class shown in the notebook. Thanks!

[cangzihan]

Maybe we can follow this demo to train a Textual Inversion model with a size of [1024]:
https://github.com/huggingface/diffusers/tree/main/examples/textual_inversion

[diptanshu-singh]

The textual inversion used in FTC seems to be different original textual_inversion, I used the textual inversion code base to learn but it would take a lot of time for each layer. Here in the assets shared, it seems like they are not learning a new token - but instead a modifier on top of layer token. I am not sure how this is being done but seems very cool thing to try out