Get the mean of a single column on the runs table

[mattiamg]

Hi. I have a project with >3K runs and I would like to obtain the Mean of a metric (e.g. accuracy, f1) across all my runs.
For example, on the runs table, I have a column “eval/f1” that contains the f1 performance on the evaluation set (at the best epoch) for each run. I would like to compute the Mean of the column which would be the Mean of the f1 performance on the development set across all runs.

I can get this number by editing the plot (Grouping) corresponding to this metric: activating Runs, grouping by None and aggregating with Mean. However, this plot can show up to the first 1000 runs (sample runs). Is there a way to obtain this plot using all 3K runs?

If not, knowing how to take the mean of the corresponding column on the Runs tables would do for my purposes.

Thanks a lot.

[uma-wandb]

Hey @mattiamg, if you’d like to find the mean value of a column for all the runs in the runs table, I recommend using our Grouping function and grouping by Grouped runs (see image), which will group by all runs in the table. From here, you can look at the column you desire for the mean (Runtime is the only column where we take the sum - we default to the mean otherwise). Please let me know if you have any questions about this!

).

[uma-wandb]

Hi Mattia, since we have not heard back from you we are going to close this request. If you would like to re-open the conversation, please let us know!