In cases where the matrix A should be is positive semidefinite, replace jnp.linalg.solve with wrapper in utils/utils.py so that we have control over the solver used.
For the time being the solver itself is not changed.
Note that the wrapper is only used is scenarios where I was reasonably confident that the matrix A is psd.
In cases where the matrix A should be is positive semidefinite, replace jnp.linalg.solve with wrapper in
utils/utils.py
so that we have control over the solver used.For the time being the solver itself is not changed.
Note that the wrapper is only used is scenarios where I was reasonably confident that the matrix A is psd.