[Bug]: `relativize_data` does not take into account sem set to None at experiment definition.

[jultou-raa]

What happened?

When utilizing an output metric with SEM set to None for inference by BOTorch, I encountered a TypeError when using the relativize option in the plot_slice_plotly function.

Error Message:

TypeError: unsupported operand type(s) for ** or pow(): 'NoneType' and 'int'

Cause: The error occurs because the relativize_data function is passing None to the relativize function within the ax.plot.helper method (specifically at line 765).

Expected Behavior: The relativize_data function should handle the None value appropriately, without resulting in a TypeError by using the inferred value.

Please provide a minimal, reproducible example of the unexpected behavior.

You can apply the example from the tutorial page by using relative=True in any function capable of relativizing to the status quo, which will implicitly invoke relativize_data.

Please paste any relevant traceback/logs produced by the example provided.

Ax Version

0.5.0

Python Version

3.12.9

Operating System

Windows

Code of Conduct

• I agree to follow Ax's Code of Conduct

[mpolson64]

Hi @jultou-raa thanks for raising this with us -- we are working on a comprehensive revamp of our plotting code and hope to get something out soon. In the meantime we can take a look at this -- would you be able to create a minimal repro which includes all the code (including experiment creation (with status quo), data attachment, and a call to the plotting function which reproduces this error?

[jultou-raa]

Hi @mpolson64

Thanks for your answer.

As I said, using tutorial page and any function with relative=True option will trigger the issue. Here is a script that can reproduce the error :

import numpy as np

from ax.service.ax_client import AxClient, ObjectiveProperties
from ax.utils.measurement.synthetic_functions import hartmann6
from ax.plot.contour import plot_contour


def evaluate(parameterization):
    x = np.array([parameterization.get(f"x{i + 1}") for i in range(6)])
    # In our case, standard error is None, since we want Ax to infer it.
    return {"hartmann6": (hartmann6(x), None), "l2norm": (np.sqrt((x**2).sum()), None)}


if __name__ == "__main__":
    ax_client = AxClient()

    ax_client.create_experiment(
        name="hartmann_test_experiment",
        parameters=[
            {
                "name": "x1",
                "type": "range",
                "bounds": [0.0, 1.0],
                "value_type": "float",  # Optional, defaults to inference from type of "bounds".
                "log_scale": False,  # Optional, defaults to False.
            },
            {
                "name": "x2",
                "type": "range",
                "bounds": [0.0, 1.0],
            },
            {
                "name": "x3",
                "type": "range",
                "bounds": [0.0, 1.0],
            },
            {
                "name": "x4",
                "type": "range",
                "bounds": [0.0, 1.0],
            },
            {
                "name": "x5",
                "type": "range",
                "bounds": [0.0, 1.0],
            },
            {
                "name": "x6",
                "type": "range",
                "bounds": [0.0, 1.0],
            },
        ],
        objectives={"hartmann6": ObjectiveProperties(minimize=True)},
        parameter_constraints=["x1 + x2 <= 2.0"],  # Optional.
        outcome_constraints=["l2norm <= 1.25"],  # Optional.
        status_quo={
            "x1": 0.5,
            "x2": 0.5,
            "x3": 0.5,
            "x4": 0.5,
            "x5": 0.5,
            "x6": 0.5,
        },
    )

    # Add status quo to the experiment and evaluate it.
    sq_trial = ax_client.experiment.new_trial()
    sq_trial.add_arm(ax_client.experiment.status_quo)
    sq_trial.mark_running(no_runner_required=True)
    ax_client.complete_trial(
        trial_index=sq_trial.index,
        raw_data=evaluate(ax_client.experiment.status_quo.parameters),
    )

    for i in range(15):
        parameterization, trial_index = ax_client.get_next_trial()
        # Local evaluation here can be replaced with deployment to external system.
        ax_client.complete_trial(
            trial_index=trial_index, raw_data=evaluate(parameterization)
        )

    plot_contour(
        ax_client.generation_strategy.model,
        metric_name="hartmann6",
        param_x="x1",
        param_y="x2",
        relative=True,
    )

[jultou-raa]

@mpolson64 were you able to reproduce this issue using the provided script ?

[esantorella]

I can reproduce this, using the latest version of Ax installed from GitHub.