-
Notifications
You must be signed in to change notification settings - Fork 16
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add expansion of params to logger #91
Conversation
@MeraX Take a look and see if this addresses the issue you faced. |
- Fix issue with duplicated keys
- Will not expand if will fit in logger
@HCookie thanks for taking this up. It looks very promising. You are expanding lists/tuples of dicts and dicts recursively. This is fine. I'm wondering if there could be any lists of lists that could be expanded as well. I think list of lists are possible in yaml, but not used in anemoi. However, I can't test your code it right now, as I have no setup ready that matches the requirements of this branch. Perhaps we could have a quick VC to look at your MLflow output? The only think I would have chosen differently is the expansion delimiter. I would prefer |
- Add tests - Alloww configuration of delim, now by default '.'
This expansion of params has been noted to cause performance issues with the mlflow servers, so may need to be considered further |
- Log all params to artifact
Is this because the number of logged parameters increases significantly? I think we saw that MLFlow struggles with this. |
@mchantry Yeah, increases from 400 -> 4000. when done naively. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks Harrison! I have left some comments and would be great to see if you have some examples from mlffow runs using this functionality! But thanks so much for the great work!
So with just logging the 'config' expanded we would get to 500? or would we still be in 1200? I believe 500 should be fine in terms of server performance |
@anaprietonem, with only |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Approving this PR as Harrison has addressed all the comments and suggestions. This changes should allow to have the parameters better logged without over-stressing the server.
…ams-string-truncation
👍 |
Closes #69
Expands iterables into dictionary of form
{key.i value_i}
for mlflow logging