Nixtla / mlforecast

Scalable machine 🤖 learning for time series forecasting.
https://nixtlaverse.nixtla.io/mlforecast
Apache License 2.0
860 stars 84 forks source link

run codspeed benchmarks in parallel #387

Closed jmoralez closed 2 months ago

jmoralez commented 2 months ago

Description

Reduces the time to run the benchmarks by using two processes to run the tests and reducing the size of the series used.

Checklist:

codspeed-hq[bot] commented 2 months ago

CodSpeed Performance Report

Merging #387 will improve performances by 40.95%

Comparing codspeed-parallel (ed86291) with main (66fee45)

Summary

âš¡ 12 improvements

Benchmarks breakdown

Benchmark main codspeed-parallel Change
âš¡ test_predict[50-1-False] 358.3 ms 254.1 ms +41.01%
âš¡ test_predict[50-1-True] 460.1 ms 309.2 ms +48.78%
âš¡ test_predict[50-2-False] 360.3 ms 255.6 ms +40.95%
âš¡ test_predict[50-2-True] 462.9 ms 310.8 ms +48.94%
âš¡ test_predict[None-1-False] 1,330.5 ms 751.5 ms +77.06%
âš¡ test_predict[None-1-True] 1,443.7 ms 818.8 ms +76.32%
âš¡ test_predict[None-2-False] 1,331.8 ms 753.6 ms +76.73%
âš¡ test_predict[None-2-True] 1,439.6 ms 821 ms +75.34%
âš¡ test_preprocess[1-False] 2.9 s 1.5 s +97.34%
âš¡ test_preprocess[1-True] 3.5 s 1.8 s +98.42%
âš¡ test_preprocess[2-False] 2.9 s 1.5 s +97.32%
âš¡ test_preprocess[2-True] 3.5 s 1.8 s +97.45%