[Feature Request] Suppress InputDataWarnings when using a non-Gaussian likelihood

[crasanders]

🚀 Feature Request

I often work with binary data that I model using SingleTaskVariationalGP with a BernoulliLikelihood. I always get a warning that says InputDataWarning: Input data is not standardized. Please consider scaling the input to zero mean and unit variance. But it doesn't make sense to scale binary data. It would be nice if this warning were suppressed when using a non-Gaussian likelihood, or if there a flag you could set to suppress it.

[esantorella]

As a quick fix, you should be able to silence that warning with from botorch import settings; settings.validate_input_scaling(False).

The suggestion sounds extremely reasonable to me. I haven't done a deep dive into the relevant code, but intuitively it seems like input checks should be based on the likelihood rather than the model type.