Streamline the logging
Right now we use a quite mixed logging system that generates plenty of csv files and matplotlib plots. It would be much more streamlined to run everything via the hydra->(optuna)->mlflow->comet-ml pipeline and to log it there. This includes the following steps:
-
Disable the matplotlib logger. -
Make all scripts independent from csv file generated by the csv logger or matplotlib logger.
In addition, this requires to re-think the testing scripts because these rely on the csv files and other logging data. See linked issue for this.
Edited by Manfred Eppe