https://kedro.org/ logo
#questions
Title
# questions
t

Tomás Rojas

03/26/2023, 9:45 PM
Hi team, how can I completely remove mlflow from a project safely?
is it enough to delete MLprojecgt file?
d

Deepyaman Datta

03/27/2023, 3:28 AM
I assume you're using
kedro-mlflow
? In which case, questions are better to also ask on their GitHub project (as it's not a plugin maintained by the core team).
y

Yetunde

03/27/2023, 8:12 AM
If it's
kedro-mlflow
then you might be able to also pop a question in #plugins-integrations
y

Yolan Honoré-Rougé

03/28/2023, 8:38 PM
Hi, I am not sure what you call "remove mlflow safely". It seems you don't use kedro-mlflow because the documentation do not nudge you to use the MLProject file, even if it can work with it. Basically it is very lilely that you use a custom implementation where a hook logs in mlflow. If it is the case, the best way is to remove the hooks in
settings.py
. You can remove the
mlruns/
folder and the
MLProject
file safely, but mlflow will automatically recreate the
mlruns/
folder at the root of the project if there is a single call to
log_param/log_metric/log_artifact/log_model
somewhere in your pipeline. I strongly recommend to centralise logging either in a custom hook or use the kedro-mlflow plugin (disclaimer: I am the author of the plugin) so you can easily switch it on / off in
settings.py
2 Views