search

Fantasy-Studio / Paint-by-Example

Paint by Example: Exemplar-based Image Editing with Diffusion Models
https://arxiv.org/abs/2211.13227
Other
1.06k stars 95 forks source link

Training Config #9

Closed JuewenPeng closed 1 year ago

JuewenPeng commented 1 year ago

Really interesting work.

I have one question about "v1.yaml" in "configs". I noticed that you set parameter "cond_stage_trainable" to "true". Does it mean that you also finetuned CLIP during training?

Looking forward to your reply.

Fantasy-Studio commented 1 year ago

We appreciate your interest in our work. We do not finetune the CLIP model. We set the parameter "cond_stage_trainable" to "true" because several MLP layers in cond_stage need to be trained and they do not belong to CLIP model.

JuewenPeng commented 1 year ago

Thank you very much.

kunalgoyal9 commented 1 year ago

@Fantasy-Studio Can we use stable-diffusion model trained using dreambooth to generate results with your pipeline?