[GENERAL SUPPORT]: where is get_optimization_trace in version 1.0.0?

[VMLC-PV]

Question

Is there any equivalent to ax_client.get_optimization_trace in v1.0.0?

I have tried the code below, but neither option worked:

trace = get_trace(client._experiment,client._experiment._optimization_config)
 or 
data = client.summarize()
trace = get_trace_by_arm_pull_from_data(data, ax_client._experiment._optimization_config)

Note that I am running a MOO.

Please provide any relevant code snippet if applicable.

Code of Conduct

• [x] I agree to follow this Ax's Code of Conduct

[Cesar-Cardoso]

Hello there! You can try the following:

from ax.analysis.plotly.arm_effects import ArmEffectsPlot

client.compute_analyses([ArmEffectsPlot(show_cumulative_best=True)])

The show_cumulative_best argument should draw a line through the best points seen so far for each trial.

[VMLC-PV]

General Kenobi, thank you! Not quite what I was looking for but still useful. The old get_trace would plot the hypervolume in case it is a multi-objective optimization, and this is what I was looking for...

[Cesar-Cardoso]

Got it! We're currently modernizing our plotting stack and that's one we haven't gotten to yet. I'm planning to work on it sometime this week. Will reference this issue in the PR.

[VMLC-PV]

Hey, Did you ever implement something on this?

[VMLC-PV]

Also, do you have any idea when the next planned release is?

[Cesar-Cardoso]

Hey! Sorry I didn't get around to implement this back when you posted it. It's in our backlog and we have it staffed, so you can expect it to be implemented in the next few months. We will update this post when we do.

As to the next Ax release, we're aiming to release version 1.1 in the coming weeks.

[VMLC-PV]

Hey, I forgot about this. Was this resolved in the latest release?