R has the mlogit library, which allows to model "discrete choice models". It accommodates both alternative (choice) specific and/or individual (chooser) specific variables. This means, e.g., if all variables are choice related, that we can limit the beta coefficients to be the same across different choices.
python has the statsmodels library. The MNLogit allows to model a "baseline-category logit model". But here it will estimate different beta's per choice. Is there any way to restrict this? Or to mix both choice and chooser specific variables?