Feature request: instruct-pix2pix

[Jonathhhan]

Would it be difficult to implement instruct-pix2pix, for example: https://huggingface.co/timbrooks/instruct-pix2pix ?
If I load the ggml model I get the error:
stable-diffusion.cpp:3045 - tensor 'model.diffusion_model.input_blocks.0.0.weight' has wrong shape in model file: got [3, 3, 8, 320], expected [3, 3, 4, 320]
I can run the inference, but the result is a black image.