For frequentist models use the vglm function in the VGAM package to run partial proportional odds models. For Bayesian partial PO models, especially constrained partial PO models, you can use the rmsb package blrm function. For vglm you can get parameter structure closer to what rms uses by specifying e.g.
f <- vglm(y ~ x1 + x2, cumulative(parallel=FALSE ~ x2, reverse=TRUE), data=...)
The notation above means to relax the PO assumption for x2.