MLXUtilsLibrary

Utilities for easing the development of machine learning inference libraries and applications on iOS and macOS using the MLX framework.

Requirements

• iOS 18.0+
• macOS 15.0+

Installation

Add MLXUtilsLibrary to your project using Swift Package Manager:

dependencies: [
    .package(url: "https://github.com/yourusername/MLXUtilsLibrary.git", from: "0.0.1")
]

Then add it to your target dependencies:

.target(
    name: "YourTarget",
    dependencies: ["MLXUtilsLibrary"]
)

Usage

Loading NumPy Files

import MLXUtilsLibrary
import MLX

// Load a single .npy file
if let arrays = NpyzReader.read(fileFromPath: npyFileURL) {
    let array = arrays["npy"]  // MLXArray
    print(array?.shape)
}

// Load a .npz archive (containing multiple arrays)
if let arrays = NpyzReader.read(fileFromPath: npzFileURL) {
    for (name, array) in arrays {
        print("\(name): \(array.shape)")
    }
}

// Load from Data
let data = try Data(contentsOf: fileURL)
let arrays = NpyzReader.read(data: data, isPacked: true)

Features

NumPy File Support

• NpyzReader: Load NumPy .npy and .npz files directly into MLX arrays
• Support for all common data types (bool, int8-64, uint8-64, float32/64)
• Handles both little-endian and big-endian formats
• Based on swift-npy

Development Tools

• BenchmarkTimer: Performance measurement utilities for profiling ML operations
• MLXArray Extensions: Enhanced debug printing for MLX arrays
• Logging Utilities: Convenient logging helpers

License

See LICENSE file for details.

Acknowledgments

• NumPy file reading based on swift-npy
• Built for use with MLX Swift