Hey Product Hunt! I'm Tudor, an independent ML engineer. I built MetalML because I wanted my everyday scikit-learn code to use my Mac's GPU without switching frameworks.

Change `import sklearn as ml` to `import metalml as ml`, keep fit, predict and transform, and supported operations run on Metal compute kernels and Metal Performance Shaders. Anything unsupported falls back to scikit-learn on the CPU, and `operation_backends_` shows you exactly what ran where.

Today it accelerates scalers, KMeans, Ridge, binary logistic regression, PCA, Naive Bayes, kNN, plus prediction for trees, forests and SVMs.

Honest status: it's experimental. GPU execution is verified on an Intel Mac with a Radeon Pro 5300M, and Apple Silicon tuning is next. If you have an M-series Mac, I'd love benchmark reports and issues on GitHub.

pip install metalml

Hey Product Hunt! I'm Tudor, an independent ML engineer. I built MetalML because I wanted my everyday scikit-learn code to use my Mac's GPU without switching frameworks.

Today it accelerates scalers, KMeans, Ridge, binary logistic regression, PCA, Naive Bayes, kNN, plus prediction for trees, forests and SVMs.

pip install metalml

MetalML

scikit-learn on your Mac's GPU, one import change

scikit-learn on your Mac's GPU, one import change