Sparse array glm fit in pymc3 very long with NUTS

so one thing I should have actually clarified, I am running this through a scipy csr_matrix in my full implenetation of it (I realise that I used np in the dummy example). your suggestion worked great on the simple example, but on my full matrix (1000 x 200000) it throws a MemoryError: None almost immediately. I assume this is a scenario where I’d need to go to AWS etc?