Skip navigation
Introducing mlx-learn: GPU-accelerated classical machine learning for Apple silicon, powered by MLX…
0:00 / 0:48

Introducing mlx-learn: GPU-accelerated classical machine learning for Apple silicon, powered by MLX…

Bojan Tunguz

@tabul_ai1.1K followers

Visit on X
200 likesDownloadOriginal post

26,644 views

Introducing mlx-learn: GPU-accelerated classical machine learning for Apple silicon, powered by MLX and native Metal kernels. Logistic regression, nearest neighbors, and exact support vector classification, with a familiar scikit-learn API. 1/4

9 Comments

Comments from the original post on X (Twitter)

  • Bojan Tunguz's profile picture

    @tabul_ai

    M4 Max benchmarks on MNIST, Fashion-MNIST, and Kannada-MNIST show up to: • 12.8× faster logistic regression training • ~11× faster KNN prediction Compared with scikit-learn, with matching or closely matching accuracy. 2/4

    11 likes

  • Bojan Tunguz's profile picture

    @tabul_ai

    SVC delivers the biggest gains in those benchmarks: • Up to 29.5× faster training with a 2 GB kernel cache • Up to 298× faster prediction Accuracy matches or closely matches scikit-learn across all three datasets. 3/4

    10 likes

  • Bojan Tunguz's profile picture

    @tabul_ai

    Use direct imports or add patch_sklearn() to accelerate supported estimators in existing workflows. Automatic CPU/GPU routing included. Available now in alpha. Open source, Apache 2.0. pip install mlx-learn Code + benchmarks: 4/4

    20 likes

  • NI's profile picture

    @pra_tha

    Congratulations!

    3 likes

  • TAAFT Research's profile picture

    @TAAFTResearch

    Congrats on shipping mlx-learn, SVC with matching accuracy at those speeds is amazing. Really congrats! Which estimators are you planning to port next after the 0.2 batch?

    2 likes

  • Kuldeep Pisda's profile picture

    @kdpisda

    was sklearn's SVC also given a 2 GB cache for the 29.5x? its default cache_size is 200MB and that alone moves libsvm timings a lot

  • Bojan Tunguz's profile picture

    @tabul_ai

    yes

    2 likes

  • Elena AI's profile picture

    @ElenaAiExpert

    Finally — scikit-learn API but actually using the Metal GPU

  • stoyan ✈️ CoRL's profile picture

    @stoyanpstoyanov

    Did you mean Canada-MNIST? Congrats on the launch!