Defining a transformation without bounds on a target should fail validation

[monken]

The NumericalTarget constructor will happily accept a transformation without any bounds. This causes the call to target.transform() to turn the dataframe into NaNs.

from baybe.targets import NumericalTarget, TargetMode, TargetTransformation

# should fail, no bounds defined
target = NumericalTarget(
    name="Target_1",
    mode=TargetMode.MIN,
    transformation=TargetTransformation.LINEAR,
)

import pandas as pd

# RuntimeWarning: invalid value encountered in divide
#   res = (upper - arr) / (upper - lower)
print(target.transform(pd.DataFrame([1, 2, 3])))

    0
0 NaN
1 NaN
2 NaN

[AVHopp]

Thanks for reporting, will have a look into this!

[AVHopp]

PR (#451 ) is up :)