how does rosetta handle hydrogens?

It depends a bit on which “mode” you’re working in.

In the most-commonly used fullatom mode (which is the default unless something explicitly switches to centroid mode), all hydrogens are present. In fact, all hydrogens are obligatory — Rosetta will automatically add in any missing hydrogens. (It will also fill in any missing heavyatoms, assuming there’s enough backbone atoms to place them.) So if you’re using standard protein residues, it really doesn’t matter if the hydrogens are present in the input structure or not. When Rosetta reads in the input, it will figure out which hydrogens are missing and add them back in. (As you’ll see in the output: all hydrogens will be present.)

Generally, you only have to worry about hydrogens when you’re making params files for ligands Rosetta doesn’t know about (e.g. for ligand docking). In those situations, you do need to have all hydrogens explicit in the SDF or MOL2 file you’re passing to the python script which makes the params file.

Centroid mode is the other “mode” to consider, used in certain protocols for a low-resoulution phase. Centroid mode uses a unified atom approach, where only polar hydrogens are represented (in addition, the protein sidechains are represented as a single superatom). But aside from that, things are similar to fullatom. Rosetta will automatically fill in any missing atoms and will ignore any extra atoms. For params file generation, the centroid parameters are typically auto-generated from the fullatom ones: either automatically internal to Rosetta, or from the Python script when the fullatom params files are first created (there’s an option to do so).