Feature request: scatter plot y=x line

[noamsiegel]

Hi

In the workspace charts, it would be helpful to plot the y=x line on a scatter plot, in a black dashed line. That would allow us to see which dots have improved upon a baseline, and which dots are worse than a baseline.

Thanks

[luis_bergua]

Hi @noamsiegel, thanks for writing in! Would you mind sharing a link to that plot so I can take a look at it and see what’s possible at this moment?

[noamsiegel]

Hi, thanks for following up.
I don’t have it anymore. It was a scatter plot in the project workspace. The annotation options were the following four, but none of them address the y=x diagonal line.

[luis_bergua]

Hi @noamsiegel, thanks for sharing this! I would recommend you to use a Weave plot as this allows you to add that line easily (here is an example). You just need to click on Add Panel → Weave and access the history of runs with runs.history.concat and then click on the panel settings to render it as a plot and customize as desired. Please let me know if this would be helpful for you!

[noamsiegel]

Thanks for the tip!

I tried that, now I see the line but I don’t see the points:
https://wandb.ai/neurohelp/chimpanzee-UNetLightningModule/reports/Personalized-Models–Vmlldzo1NTgyNTUy#section-1

And in the report you will see the original scatter plot for reference:

[luis_bergua]

I see, thank you so much for sharing this @noamsiegel! I just took a look at this, that behavior is because those metrics are logged separately so the history contains them in different rows as you can see in the following images:

You can use a Weave expression like this [{"baseline/false_alarms_in_24_hours": runs.history.concat["baseline/false_alarms_in_24_hours"].dropna, "enhanced/false_alarms_in_24_hours": runs.history.concat["enhanced/false_alarms_in_24_hours"].dropna}] to remove nulls from both metrics and then create a new table and plot it, this will show your points

Let me know if that helps!

[luis_bergua]

Hey @noamsiegel, just wanted to follow up here and ask if the solution I proposed would be helpful for you. Please let me know if I can help you in any other way!

[luis_bergua]

Hi @noamsiegel , 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!