It was recently found out that the jax.experimental.sparse module will now only be supported with no more active development found here. Since we are solving sparse systems throughout this workflow, a change will most likely be needed in the near future to pivot from this dependency. We are implicitly dependent on this library through the sparsify operation occurring in the SparseMatrixLinearOperator.
There are two potential options to pivot towards: PETSc and CuPy. We should weigh the options between the two before making a decision. This issue is created to layout the thoughts behind which we will choose in the coming future. Most likely will involve a simple implementation of both and see which one will pass a suite of test cases.
It was recently found out that the jax.experimental.sparse module will now only be supported with no more active development found here. Since we are solving sparse systems throughout this workflow, a change will most likely be needed in the near future to pivot from this dependency. We are implicitly dependent on this library through the sparsify operation occurring in the SparseMatrixLinearOperator.
There are two potential options to pivot towards: PETSc and CuPy. We should weigh the options between the two before making a decision. This issue is created to layout the thoughts behind which we will choose in the coming future. Most likely will involve a simple implementation of both and see which one will pass a suite of test cases.