PyCoreML

Run CoreML models from Python — load .mlmodel files, make predictions, and control compute units.

Install

pip install pycoreml

API

Function	Description
`load_model(path)` → `model`	Load a `.mlmodel` or compiled `.mlmodelc`
`predict(model, inputs)` → `dict`	Run inference with input dictionary
`model_description(model)` → `str`	Get model metadata and description
`set_compute_units(units)`	Set compute preference: `"all"`, `"cpu"`, `"gpu"`, `"neural_engine"`

Usage

import pycoreml as ml

# Load a model
model = ml.load_model("MyClassifier.mlmodel")

# Get model info
print(ml.model_description(model))

# Run prediction
result = ml.predict(model, {
    "input_feature": 42.0,
    "image": [0.1, 0.2, 0.3, ...],
})
print(result)

# Control compute units
ml.set_compute_units("neural_engine")  # Use Apple Neural Engine
model = ml.load_model("MyModel.mlmodel")
result = ml.predict(model, {"x": 1.0})

Compute Units

Value	Description
`"all"`	Auto-select best (default)
`"cpu"`	CPU only
`"gpu"`	GPU (Metal)
`"neural_engine"`	Apple Neural Engine