/modelserver

creates a small server (plumber) that tracks your modelrun properties. Like mlflow tracking,but lighter weight

Primary LanguageR

A server for machine learning analytics.

Active Repository licence MIT The idea is you send a request when you start a run and when you end a run and in between you send metrics to the server.

The server handles these things and puts them into a database. (this is inspired by the MLflow python package that does this and more) I think this is most like MLFlow - Tracking.

You can find this project on gitlab and github.

How to run

Right now, clone this repo somewhere. In the commandline go to this somewhere. Start R in that location. run renv::restore() move out of R. In the terminal run the command Rscript run_server.R

Comparison with MLFlow

MLFlow uses runs and experiments. A run is an execution of code. multple runs live under 1 experiment. A run also includes timing.

wow they improved the docs a lot in the past years! I love it, it is much easeir to understand.

So I did the same thing, but named it differently. I have ids for runs and optional group to combine several.