Hi kedro users, is anyone familiar with how we can toggle

kedro-mlflow

on/off in the local config? I understand that we can deactivate tracking for certain pipelines, but on some runs I don't want MLFlow to run at all.

Hi Kok, you can disable plugin's hooks in the

settings.py

DISABLE_HOOKS_FOR_PLUGINS = ("kedro-mlflow", "kedro_mlflow")

@Takieddine Kadiri Thank you, but is there any way I can set this as a config, e.g. in

conf/local

?

I'd love to, but AFAIK this does not exist in mlflow. If mlflow were to expose an environment variable, I could set it in the cinf but this does not exist! (I think it would be a good issue for mlflow itself)

@Yolan Honoré-Rougé maybe this: https://github.com/Galileo-Galilei/kedro-mlflow/blob/dc3b7bf98a50f1dc5048b85e42a2620bde1de466/kedro_mlflow/framework/hooks/mlflow_hook.py#L22[…]3C32 Should be transformed into a method that will allow users to override the check in their own hooks that will inherit from

MLflowHook

? 🤔

Not sure of what you want to do @marrrcin. I think the OP wants something like

kedro run --no--mlflow

but this is not easy to achieve because kedro does not enable to supercharge plugin this way (there's an issue about it but I could eventually hack it). The real reason I can't do it is because there is no easy way to "deactivate" mlflow. If anyone has a

mlflow.log_xxx

call inside a node, this will create a mlrun folder anywhere and the plugin will not know about it. The pipeline deactivation feature is just me hacking datasets and hooks to prevent logging, but this is not really a deactivation of mlflow