-
Notifications
You must be signed in to change notification settings - Fork 22
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
fix: Remove the arrow up and down from the column name in the dataframe #1260
Comments
From a discussion with @adrinjalali IRL, we could make the metric being the row instead of column of the dataframe. It allows 2 things:
Transposing makes sense because right now, if one want to plot with pandas, you need to write: report.metrics.report_metrics().T.plot.barh() We can simplify it to report.metrics.report_metrics().plot.barh() which make total sense to me. |
An additional thought is about polars/pandas dataframe. Currently, we have multindex by default (because I like them). I think that we all agree that multindex are nice when we look at the HTML representation but they are not easy to deal with when indexing. Therefore, I think that we need a parameter to explicitely request flat vs. multindex. For instance, in the model cards, we want only the representation and multindex make sense. When a user call |
Thanks, +1 for "multi index are nice for visualization but not easy for indexing" |
I also think that it should be in the 0.7 milestone. Once we transpose, it will be pretty straightforward. ping @MarieS-WiMLDS |
Same as the other issue: if it's ready, it's a bonus! It will be at least necessary for 0.8 where we want to include improvement about the reports. |
closes #1260 Add whether a metric is greater is better or lower is better as a column indicator instead to have it in the name of the index.
When showing a score dataframe, e.g.
report.metrics.report_metrics()
, the column name contains some arrow up/down depending if we show a score or a loss. However, adding this information directly in the column is annoying to index a specific score.We should come with another way to provide this information. Maybe, a dirty way but possible way would be to patch the HTML repr from pandas for those specific dataset to incorporate the information in the HTML representation itself.
The text was updated successfully, but these errors were encountered: