search

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

Question about IP-Adapter FaceID plus #326

Closed vim-hjk closed 6 months ago

vim-hjk commented 6 months ago

Hello, I have a few questions. Are LoRAs corresponding to ["ip_adapter"] in ip-adapter-faceid-plusv2_sd15.bin and LoRAs corresponding to ip-adapter-faceid-plusv2_sd15_lora.safetensors different LoRAs?

When I check the weight tensor value, it looks different. If it is different, can I know how to apply LoRAs corresponding to ip-adapter-faceid-plusv2_sd15_lora.safetensors during training?

Thank you.

xiaohu2015 commented 6 months ago

it should be same, the ip-adapter-faceid-plusv2_sd15_lora.safetensors is converted from ip-adapter-faceid-plusv2_sd15.bin