Can I input discrete sets for Bayesian optimization?

Hello, everyone. I am a student studying machine learning. I’m trying to find the best parameter sets using Bayesian optimization(The framework is Botorch).

For example, I enter this training data :

input_A=[[0., 50., 100., 150., 0., 100.],
       [50., 50., 0., 150., 150., 100.]]

output_A=[[284.53],
        [285.62]]

The more I iterate, the less new parameter sets come out.

To summarize my question,

  1. Can the optimization proceed properly even if a tensor consisting of 0, 50, 100, and 150 is used like input_A?

  2. I changed the parameter sets obtained from the acquisition function through rounding to 0, 50, 100, 150. Can I make 0, 50, 100, 150 come out in a way that does not round the parameter sets obtained?

Thank you for read.