Giters

allegroai / clearml

ClearML - Auto-Magical CI/CD to streamline your AI workload. Experiment Management, Data Management, Pipeline, Orchestration, Scheduling & Serving in one MLOps/LLMOps solution

Home Page:https://clear.ml/docs

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Color selection in Reports

gabmis opened this issue · comments

commented

Hi! Thank you for the great work!

Proposal Summary

Reports are very nice but would be significantly better if we were able to modify plot colors in them.

Motivation

It's useful to have visible (e.g. in dark mode) and coherent (throughout the report) colors.
Currently, we can't seem to have that because we can't set colors in reports and the colors we set in the experiment comparison view aren't persistent w.r.t the embedding operation.

Related Discussion

Couldn't find any related discussions

Best,
Gabriel Misrachi

commented

Thanks for pointing this out @gabmis.

Currently, we can't seem to have that

Since plot colors persist across comparisons (i.e. the same experiment will have the same color), and embedding a plot basically retrieves those plots - The solution would be to make sure that the color transformation into reports (to keep dark mode visibility) is coherent rather than going and manually changing the color for all plots manually. Wouldn't you agree?

commented

Hi! thanks for answering so quickly, my bad for taking so much time to respond (was off of work).
Yes that sounds like a desirable solution. Are you considering implementing it?
By the way, although I had observed it myself, I could not find any documentation to ensure this persistence (between comparison view) existed, do you know if there is any.
Best,
Gabriel

commented

@gabmis the color dark-mode transformation persistency issue should be fixed in future versions.

Though not explicitly documented, this is a core principle of the ClearML UI, that a plot variable should use the same color across plots (unless manually overridden). This should hold for all plots, and specifically in comparison (where the "variable" is the experiment name). What documentation were you looking for?

commented

I was not aware of that principle. That's what I was looking for.

Not sure we're understanding the same thing.
I thought your point was that colors which are set in the comparison view should persis into the reports. Is that what you're saying?