Initialize UNet weights

[gaoyixuan111]

Hello, may I ask if you did not add the corresponding layer weights when initializing the UNet model in the training code? For example, "attn_procs[name].load_state_dict(weights, strict=False)". @RedAIGC

[RedAIGC]

Hello, may I ask if you did not add the corresponding layer weights when initializing the UNet model in the training code? For example, "attn_procs[name].load_state_dict(weights, strict=False)". @RedAIGC

Thanks for your interest, we initial the lora weights with state_dict from ip-adapter-faceid-plusv2_sdxl.bin