coreml

Load and run Core ML models on-device with Swift APIs.

Updated Apr 19, 2026

Search Tags:#swift #vision #on-device #machine-learning #coreml #mlmodel #compute-plan

Ready to run

npx skills add https://github.com/femitz/flyby --skill coreml-femitz

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: coreml
Source: https://github.com/femitz/flyby/tree/main/.agents/skills/coreml
Command: npx skills add https://github.com/femitz/flyby --skill coreml-femitz

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Load and run Core ML models on-device through Swift APIs and configuration.

Core Features & Use Cases

Auto-generated Class Usage: Use Xcode-generated model classes for typed inputs and outputs during predictions.
Manual MLFeatureProvider: Build custom feature providers for dynamic input schemas.
Vision/NLP Integration: Leverage Vision or Natural Language for preprocessing and results when needed.

Quick Start

Load a model, configure compute units, and run a prediction to start on-device ML.

Frequently Asked Questions about coreml

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I load and run a Core ML model on-device using Swift?▼

To run a Core ML model on-device, load the compiled .mlmodelc or .mlpackage file using Swift APIs, configure the compute units via MLModelConfiguration, and execute predictions either through Xcode auto-generated classes or manual MLFeatureProvider inputs.

What is the difference between using auto-generated classes and MLFeatureProvider for predictions?▼

Auto-generated classes provide typed inputs and outputs for static model schemas, while manual MLFeatureProvider construction allows you to build custom feature providers for dynamic or flexible input schemas during on-device predictions.

How do I configure compute units for an on-device machine learning model?▼

Compute units are configured using MLModelConfiguration in Swift, allowing you to assign model execution to the CPU, GPU, or Neural Engine before loading the model and running predictions on-device.

Can I use Vision or Natural Language frameworks for preprocessing Core ML inputs?▼

Yes, you can leverage Vision or Natural Language frameworks to handle image or text preprocessing and parse prediction results before or after running the Core ML model on-device.

Do I need to convert my model to .mlpackage or .mlmodelc to run predictions?▼

Core ML requires models in the compiled .mlmodelc format to execute predictions on-device, though .mlpackage formats are also supported for loading and running workflows through Swift APIs.

How do I handle MLMultiArray data for Core ML model predictions?▼

MLMultiArray handling is required for managing multi-dimensional data structures when passing inputs to and extracting outputs from Core ML models via MLFeatureProvider or auto-generated classes.