Multivariate distribution

I noticed that in scipy, the multivariate distribution is always parameterized by the mean vector and the covariance matrix. I have a problem where parameterization with a sparse precision matrix (inverse of the covariance matrix) is more appropriate.

Is there any class that would allow me to parameterize it directly with a precision matrix? The number of variables is quite large and I would like, if possible, to avoid calculating the inverse of this sparse precision matrix.