pymlbenchmark: benchmark around machine learning

This project started with my first attempt to bring a modification to scikit-learn The template was reused to measure various implementations or models and to make them shorter but with probably less understanding of the whole script. It evolved into providing classes to compare ONNX runtimes such as onnxruntime to scikit-learn.

• GitHub/pymlbenchmark

• documentation

• Blog

current - 2021-01-05 - 0.00Mb

• 12: Support double when converting to onnx (2019-12-21)

0.2.141 - 2019-12-13 - 0.04Mb

• 11: Fix dimension when calling convert_sklearn (2019-10-07)

• 8: implements OnnxRuntimeBenchPerfTestRegression (2019-07-15)

• 6: use assume_finite=True whever possible (2019-07-15)

• 4: update colors in graph, use similar colors but not equal (2019-05-23)

• 5: run profiler on some calls and saves the results (2019-04-23)

• 3: add mechanism to dump error, data to investigate later when it happens (2019-03-12)

• 1: move benchmark from _benchmarks (2019-03-12)

0.1.30 - 2019-03-07 - 0.02Mb

• 2: first commit (2019-03-05)