How does pymc3 handle n=0 in Binomial discrete likelihood function

Thanks for noting how I can check it.
Now I can check the arrays of trials and observations with the zeros, shortened without the zeros, and with NaN.