Avoiding memory error when using scipy.sparse.block_diag

Question

I have the following code for creating a very large Jacobian, where I first define a Jacobian with respect to one third of the relevant vector, before this is then converted