Generated Images are 512x512 using stable-diffusion-v1-4, how can I generate higher res?

[facinick]

[3DTOPO]

We have super resolution running on device. Does an excellent job of enlarging - reduces noise, sharper lines while increasing resolution.

[3DTOPO]

Also SD 2 has weights trained at 768, so I suppose you could use that model and increase the output shape to 768 in the convert to coreml script. But, that means inference time and RAM requirements will double.

[dreampuf]

Hi @3DTOPO, have you managed to use huggingface.co/stabilityai/stable-diffusion-2? I got a runtime error below.

Traceback (most recent call last):
  File "/opt/homebrew/anaconda3/envs/coreml_stable_diffusion/lib/python3.8/runpy.py", line 194, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/opt/homebrew/anaconda3/envs/coreml_stable_diffusion/lib/python3.8/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/Users/coder/hub/ml-stable-diffusion/python_coreml_stable_diffusion/pipeline.py", line 534, in <module>
    main(args)
  File "/Users/coder/hub/ml-stable-diffusion/python_coreml_stable_diffusion/pipeline.py", line 478, in main
    image = coreml_pipe(
  File "/Users/coder/hub/ml-stable-diffusion/python_coreml_stable_diffusion/pipeline.py", line 354, in __call__
    image = self.decode_latents(latents)
  File "/Users/coder/hub/ml-stable-diffusion/python_coreml_stable_diffusion/pipeline.py", line 200, in decode_latents
    image = self.vae_decoder(z=latents.astype(np.float16))["image"]
  File "/Users/coder/hub/ml-stable-diffusion/python_coreml_stable_diffusion/coreml_model.py", line 79, in __call__
    return self.model.predict(kwargs)
  File "/opt/homebrew/anaconda3/envs/coreml_stable_diffusion/lib/python3.8/site-packages/coremltools/models/model.py", line 509, in predict
    return self.__proxy__.predict(data)
RuntimeError: {
    NSLocalizedDescription = "Error computing NN outputs.";
}

[3DTOPO]

It is running at the moment - on my Ubuntu farm.

I was interested in possibly using it for apps in development but for my purposes I think it's a bit early so have given up on the idea for now.

[brunoamaral]

I got exactly the same error as @dreampuf trying to run SD V2 on Python 3.10.6

[blu-Fox]

I got the same error when I used a CoreML model I converted myself. When I used a pre-converted model from Apple's repo on HuggingFace, SD2base successfully ran in Python using Terminal. I don't know how to make SD2.1 work, how to increase resolution to 768, or how to upscale