topic Re: Avoid Using MLFlow to log runs in Machine Learning

Avoid Using MLFlow to log runs

Evan_MCK — Fri, 10 Feb 2023 17:48:40 GMT

I am running some code to smooth columns of data and fill in missing values. The algorithms being used are automatically logged to an experiment using MLFlow which makes a 15 second job take 9 minutes and causes out of memory errors in another instance. I have not even imported the MLFlow package in the notebook. How do I avoid logging results?

Re: Avoid Using MLFlow to log runs

Evan_MCK — Fri, 10 Feb 2023 20:06:07 GMT

I think I figured it out. Autologing must have been enabled from a previous run. Pretty easy to solve. Posting this to help anyone else in this situation.

https://mlflow.org/docs/latest/tracking.html#automatic-logging

To disable just run the appropriate command for the library being logged as per below.

import mlflow

mlflow.sklearn.autolog(disable=True)

mlflow.xgboost.autolog(disable=True)

mlflow.statsmodels.autolog(disable=True)

Re: Avoid Using MLFlow to log runs

Evan_MCK — Wed, 28 Jun 2023 14:32:54 GMT

You can even us a general command without noting which library will be used:

import mlflow

mlflow.autolog(disable=True)

Re: Avoid Using MLFlow to log runs

mgorjis — Wed, 17 Sep 2025 23:55:26 GMT

very helpful. WIthout this command, optuna tries to register the info of each experiment, which make it so slow.