Question about seasonality variables

Hello!

I recently spent a little time trying to hyperparameter tune a model on a dataset and I came across a slightly conflicting bit of documentation. In the docs on tuning (https://facebook.github.io/prophet/docs/diagnostics.html#hyperparameter-tuning), there are 3 (yearly/weekly/daily)_seasonality parameters that are documented under the section 'probably won't be tuned', with accepted values of 'auto' / True / False .

Both the implementation and the paper suggest that these parameters can also be an integer that can be tuned (beyond the defaults that 'auto' provides). For my dataset, the cross validated performance metrics did improve as I tuned these. Should tuning these be avoided?

Thanks!