nevergrad_pso fails when bounds are not provided

[spline2hg]

Description:
I am trying to use the nevergrad_pso optimizer without specifying bounds, as it has needs_bounds=False. However, I encounter a ValueError requiring at least one bound to be specified.

To Reproduce:

import optimagic as om
import numpy as np

def sphere(params):
    return params @ params

res = om.minimize(
    fun=sphere,
    params=np.arange(5),
    algorithm="nevergrad_pso",
)

Error messages:

    140                 raise ValueError(f"Lower bounds {a_min} should be strictly smaller than upper bounds {a_max}")
    141         if self.a_min is None and self.a_max is None:
--> 142             raise ValueError("At least one bound must be specified")
    143         self.shape: tp.Tuple[int, ...] = self.a_min.shape if self.a_min is not None else self.a_max.shape
    144 

ValueError: At least one bound must be specified

I am not very sure if this is a bug ,or am I missing something?

[janosg]

@gauravmanmode can you look into this?

[gauravmanmode]

Sorry for the confusion.
Earlier we used set_bounds for setting bounds when bounds were required. This is fixed by settings bounds through ng.p.Array constructor which has optional bounds.

• This is fixed in the PR Migrate nevergrad optimizers to new documentation style #632 and I have also fixed the test.

[janosg]

@spline2hg I close the issue then.