ML on a Mac

Steps for Machine Learning on Apple Silicon M1/M2 chips, with Stable Diffusion

Jan 23, 2023 2 min read

Stable Diffusion converted to Core ML by Apple

You can use Apple’s ml-stable-diffusion to get started or try my fork with a script for convenience.

Feel free to git clone my fork, or use Apple’s repo, and then run that script.

So far Apple has only converted a few models to Core ML. Check out to find a lot more models including Stable Diffusion 2.1.

Download models for Stable Diffusion

Modify scripts/ (source) to choose the model you’d like to download. Then run it from the command line: python scripts/

Check the models/<output> folder to see the model. If it is a model (models converted to Core ML), then unzip it first.

Run inference for Stable Diffusion

After downloading the model using the script above, run this in the command line to generate images for a given prompt:

# MODEL=coreml-stable-diffusion-v1-4_original_compiled
# MODEL=coreml-stable-diffusion-v1-5_original_compiled
# MODEL=coreml-stable-diffusion-2-1-base_split_einsum
# COMPUTE_UNITS=all      # "split_einsum" models
COMPUTE_UNITS=cpuAndGPU  # "original" models
mkdir -p $OUTPUT_PATH

PROMPT="a photograph of an astronaut riding on a horse"
SEED=42 # 93 is the default
echo "Generating \"$PROMPT\" on $MODEL with seed $SEED"
time swift run StableDiffusionSample $PROMPT --resource-path models/$MODEL --compute-units $COMPUTE_UNITS --output-path $OUTPUT_PATH --seed $SEED


Read more posts like this in the Software Engineering Toolbox collection.
Visit homepage
comments powered by Disqus