Initialize UNet weights

RedAIGC / StoryMaker

StoryMaker: Towards consistent characters in text-to-image generation

575 stars 50 forks source link

Initialize UNet weights #33

Open gaoyixuan111 opened 2 weeks ago

gaoyixuan111 commented 2 weeks ago

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