When running a lot of experiments where we are only interested in the pipeline log, it can help to have an option which cleans up all information from the database and temporary files (e.g., models) to save disk space. For example, when running Criteo experiments, it sometimes happened that the huge DLRM snapshots filled up the disk.
When running a lot of experiments where we are only interested in the pipeline log, it can help to have an option which cleans up all information from the database and temporary files (e.g., models) to save disk space. For example, when running Criteo experiments, it sometimes happened that the huge DLRM snapshots filled up the disk.