A Guide to Implementing MLOps by Prafful Mishra

Author:Prafful Mishra
Language: eng
Format: epub, pdf
ISBN: 9783031820106
Publisher: Springer Nature Switzerland

What ML-specific logs must have for a production ready application?

There are certain aspects that need to be logged for ML systems, such as the dataset name and version used, model name and version used, hyperparameters used, time taken to infer a request, etc.

Should there be a common convention for logging across the organization?

Yes, as mentioned above this can be very valuable in case of on-call issues, as the person looking at the logs has their eyes already oriented to look at one specific structure of the logs, which would directly relate to the time taken by them to understand the root cause of the problem.

This can also enable cross team collaboration, as it becomes easier for people from other parts of the organization to understand what is happening with applications that they do not necessarily maintain.

There might obviously be certain settings that might only make sense for ML specific applications/teams. This would mean that teams/functions could have their own extended implementation of the base logging module. Something like:

from myorg.logging import logger # for normal logging

from myorg.logging.ml import logger # for ML specific applications

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.