Core ML will try to run your model on the ANE whenever possible, but you can't force Core ML to use the ANE.
When creating an instance of your Core ML model, you can pass in an MLModelConfiguration object:
let config = MLModelConfiguration()
config.computeUnits = .all
let model = try MyModel(configuration: config)To allow the model to run on the ANE, use computeUnits = .all.
Note that this does not guarantee the model will run on the ANE, it just tells Core ML that you prefer to use the ANE if available. If the device does not have an ANE, Core ML will automatically fall back to the GPU or CPU.
If possible, Core ML will run the entire model on the ANE. However, it will switch to another processor when it encounters an unsupported layer. Even if Core ML could theoretically run the second part of the model on the GPU, it might actually decide to use the CPU. You can't assume anything here. (Note that the CPU can actually be faster than the GPU on many neural network operations, so it's not necessarily a bad thing.)