How to Specify Shortcut Scale Parameters when training

tencent-ailab / IP-Adapter

The image prompt adapter is designed to enable a pretrained text-to-image diffusion model to generate images with image prompt.

Apache License 2.0

5.08k stars 331 forks source link

How to Specify Shortcut Scale Parameters when training #259

Open Ted-developer opened 8 months ago

Ted-developer commented 8 months ago

I have noticed that FaceID Plus is differentiated between v1 and v2. Could you please confirm if the v2 training includes an added shortcut parameter toggle?

When training the FaceID Plus v2 model with "shortcut = True," how should the "scale" parameter be specified?

Thank you very much.

xiaohu2015 commented 8 months ago

Thank you for your attention. There are too many recent news to reply to one by one. You may be interested in how the face model is trained. I will write a detailed article later to explain it.