Skip to main content

Matft

Matft is a Numpy-like multi-dimensional array library in Swift. Function names, arguments and behaviors follow Numpy (e.g. np.expand_dims → Matft.expand_dims), so you can port Python code almost line by line.

import Matft

let a = Matft.arange(start: 0, to: 27, by: 1, shape: [3, 3, 3])
let b = a[1~<3, Matft.all, 0]          // a[1:3, :, 0] in Numpy
let c = Matft.math.sin(a) + a.T        // broadcasting & element-wise math

Features​

  • MfArray — an n-dimensional array like numpy.ndarray, with many types (Bool, UInt8…Int64, Float, Double, complex) and pretty print.
  • Indexing — positive / negative indexing, slicing (~<), newaxis, boolean and fancy indexing.
  • Views — slicing returns a view that shares memory with the original array, as in Numpy.
  • Broadcasting, transpose, reshape, astype and universal function reduction.
  • Math, statistics, linear algebra, FFT backed by Apple's Accelerate framework.
  • Complex numbers.
  • Image processing compatible with OpenCV (cv2), and preprocessing compatible with PIL / Hugging Face transformers.
  • Audio features (STFT, mel spectrogram, Whisper log-mel) compatible with librosa.
  • MLX interop — zero-copy conversion between MfArray and MLXArray via the separate MatftMLX package.

Matft is a complement to mlx-swift: Matft plays the role of NumPy / SciPy / OpenCV / librosa, and MLX plays the role of PyTorch. See Matft and MLX for the comparison.

Where to go next​

  • Installation and Quick Start
  • Guide — how to use MfArray step by step
  • NumPy Mapping — the list of Matft functions and their Numpy / OpenCV / librosa counterparts
  • API Reference — every public symbol, generated from the source by Swift-DocC (see the navbar)