← ClaudeAtlas

coremllisted

Integrates and profiles Core ML models for on-device inference. Use for mlmodel/mlpackage loading, generated or feature-provider predictions, compute-unit selection, MLTensor, Vision integration, MLComputePlan, model pipelines, deployment, or performance analysis.
thienqo/ios-skills · ★ 11 · AI & Automation · score 69
Install: claude install-skill thienqo/ios-skills
# Core ML Swift Integration Load, configure, and run Core ML models in iOS apps. Covers Swift model loading, synchronous/async inference, batch prediction, MLTensor, profiling, and memory management. > **Scope boundary:** Python-side model conversion, quantization, and pruning belong in `apple-on-device-ai`. This skill owns Swift-side integration only. ## Contents - [Loading Models](#loading-models) - [Model Configuration](#model-configuration) - [Making Predictions](#making-predictions) - [MLTensor (iOS 18+)](#mltensor-ios-18) - [Vision Integration](#vision-integration) - [Performance & Memory](#performance--memory) - [Common Mistakes](#common-mistakes) - [Review Checklist](#review-checklist) - [References](#references) ## Loading Models - **Auto-Generated Class**: Add `.mlmodel` or `.mlpackage` to the target; Xcode generates typed input/output classes. - **Async Loading (iOS 15+)**: Avoid blocking the main thread: `try await MLModel.load(contentsOf: url, configuration: config)`. - **Runtime Compilation (iOS 16+)**: Compile downloaded packages with `MLModel.compileModel(at: url)`. Cache the resulting `.mlmodelc` URL in Application Support; recompiling on every launch is an error. ```swift import CoreML let config = MLModelConfiguration() config.computeUnits = .all // Typed model initialization let classifier = try MyImageClassifier(configuration: config) // Dynamic async loading let model = try await MLModel.load(contentsOf: modelURL, configuration: config) ``` ##