First, install maria if you have not already.  Following https://thomaswmorris.com/maria/installation.html, I would recommend installing from source as:

git clone https://github.com/thomaswmorris/maria
cd maria
pip install -e .

The online tutorial on custom map simulations (https://thomaswmorris.com/maria/tutorials/custom-map-simulations.html) shows how to set up a 2-color (dichroic) camera in maria and perform a basic simulation.  However, that does not include polarization; this one shows how to do that: https://thomaswmorris.com/maria/tutorials/polarized-observations.html.

Here in this folder, the files atlast-camera-concepts.pdf (PDF version) or atlast-camera-concepts.ipynb (Python Jupyter notebook) show in general how to define 9 bands, corresponding to those in the AtLAST SZ case, that could fill large portions of the FoV.  Beware, the detector counts grow into the millions for this case, so the notebook is largely illustrative.

The files atlast_9band_cameras.pdf (PDF version) or atlast_9band_cameras.ipynb (Python Jupyter notebook) show what's in the atlast.yml files now in maria (maria/bands/configs and maria/instrument/configs). 

The python scripts atlast_sz_mini_9band_maria_sim_cl00003_1hr.py and atlast_toltec-like_maria_sim_cl00003_1hr.py show 1 hour observation examples with the "mini" version of the 9-band camera and with a TolTEC-like 3-band instrument on AtLAST.
atlast_sz_mini_9band_maria_sim_cl00003_1hr.py has more detailed annotations.

Notably, the maximum likelihood mapmaker (see https://thomaswmorris.com/maria/tutorials/maximum-likelihood-mapper.html) results in much nicer maps right away, but I've found it doesn't run for me on my laptop (miniconda on an Apple Silicon Mac), but it runs well on my local Linux cluster.
