macOS Machine learning resources on Apple Silicon

[Cromulent]

I have an M1 Mac Studio and know they come with a 16-core Neural Engine. I've been interested in machine learning but haven't seen any Apple-focused information on how to use the hardware in Apple Silicon macs.

Does anyone have any good resources for A) machine learning foundations and theory and B) for using the Neural Engines in Apple Silicon macs in Swift?

 

[apostolosdt]

Recently, I was looking for any intro info and I came across those two sites:

https://kazmer-nagy-betegh.com/post/m1-machine-learning-setup/

https://www.anaconda.com/blog/apple-silicon-transition

which I hope you will find useful (sorry if already known). Regarding books, I am aware of the following:

Hastie et al., The Elements of Statistical Learning, 2nd ed., Springer;

a highly recommended text.

 

[casperes1996]

To utilise the Neural Engine you just want to be running models on CoreML. You can do simple training with the CoreML app on macOS to create a model, though I believe you can also convert a TensorFlow or PyTorch model over to CoreML to use with the CoreML framework for utilising the Neural Engine.
ML is not really my field in general so I can't say much more specifically about this, I've only fooled around with very basic things, but for using the NE, CoreML is your answer. Some people have had success sending more general commands to it, but it is not supported as a generic computation device and generally you access it only through Apple frameworks and CoreML.

 

[Cromulent]

Recently, I was looking for any intro info and I came across those two sites:

https://kazmer-nagy-betegh.com/post/m1-machine-learning-setup/

https://www.anaconda.com/blog/apple-silicon-transition

which I hope you will find useful (sorry if already known). Regarding books, I am aware of the following:

Hastie et al., The Elements of Statistical Learning, 2nd ed., Springer;

a highly recommended text.

To utilise the Neural Engine you just want to be running models on CoreML. You can do simple training with the CoreML app on macOS to create a model, though I believe you can also convert a TensorFlow or PyTorch model over to CoreML to use with the CoreML framework for utilising the Neural Engine.
ML is not really my field in general so I can't say much more specifically about this, I've only fooled around with very basic things, but for using the NE, CoreML is your answer. Some people have had success sending more general commands to it, but it is not supported as a generic computation device and generally you access it only through Apple frameworks and CoreML.

Thank you both. Very interesting information.