erasuer-tokens for other dataset

[jsw6872]

Thank you for impressed studies.

It's same question on title.
Is there any erasure-tokens for any other dataset? (ex. flowers102, imagenet, coco, caltech101)
or how to get erasure-tokens weight for custom dataset ?

[brandontrabucco]

Hello jsw6872,

Thanks for your interest in our work! Yes, I'll upload and share these in a google drive link shortly.

There are: (1) SD checkpoints fine-tuned to erase concepts, and (2) tokens specific to these SD checkpoints for DA-Fusion.

I will upload and share both for the missing datasets.

[jsw6872]

Hello jsw6872,

Thanks for your interest in our work! Yes, I'll upload and share these in a google drive link shortly.

There are: (1) SD checkpoints fine-tuned to erase concepts, and (2) tokens specific to these SD checkpoints for DA-Fusion.

I will upload and share both for the missing datasets.

Thank you for reply :)
If you'll excuse me, I have one more question for you.

I see that the embed token is required when generating synthetic images with textual inversion, how to generate a embeding .pt file for a specific dataset? Is it right fine_tune.py? I want to make flowers101 dataset embedding tokens..