Hello,
Just wanted to check with you if this is what you wanted to do, here in the denoising loop you only use 2 block outputs while there are 3, and how you trained the layers controlnet_blocks.2.weight and controlnet_blocks.2.bias if they are unused ?
Thanks