I have a multiple runs with a varying hyperparameter (for simplicities sake: number of training samples).
I would like to plot the mean and standard deviation of the accuracy after finished training in relationship to the number of training samples used.
y=accuracy
^
| ---------- x
| ------ x
| --x
_____________ x-axis = training samples from config
Currently I only see wandb reports being able to plot metrics against their time/step stamp.
Weave is a possibility but even afters years of matplotlib use I haven’t figured out how to use weave as I can’t find a understandable documentation.
Exporting the data to a Jupyter notebook is also a possibility which I have done, but it sort of undermines the purpose of wandb for quick run comparison and visualization as it includes laborious fiddling with matplotlib.
Any help on how to use config values on the x-axis of plots is greatly appreciated.
Hi @ludwigwinkler thanks for writing in! Since you have multiple runs, you will have to add the panel in the Project level. You could add a Scatter plot and plot the config variable against accuracy. Have you logged the mean and std of the accuracy? If you edit the Panel, in the Annotations tab, there are some running aggregate metrics. Would these work for your use case?
Hi @ludwigwinkler I wanted to follow up with you on this request, and see if the above option would work for you, or if you were looking into a different solution such as Weave? Thank you!
Hi @ludwigwinkler since we haven’t heard back from you, I will go ahead and close this ticket for now. Please feel free to message us back if you’re still experiencing any issue, and we will be glad to keep investigating!
