Apparently prior releases of openfermion were attempting to
use scipy.linalg.expm to exponentiate sparse matrices. Until recently
the scipy API seem to resolve the issue of using scipy.linalg.expm
instead of scipy.sparse.linalg.expm.
This issue was not surfaced until the 1.0.0dev release so some
combination of numpy/scipy pinning or depinning has revealed this
bug in openfermion. This commit fixes the issue by converting
scipy.linalg.expm to scipy.sparse.linalg.expm where appropriate.
Apparently prior releases of openfermion were attempting to use scipy.linalg.expm to exponentiate sparse matrices. Until recently the scipy API seem to resolve the issue of using scipy.linalg.expm instead of scipy.sparse.linalg.expm.
This issue was not surfaced until the 1.0.0dev release so some combination of numpy/scipy pinning or depinning has revealed this bug in openfermion. This commit fixes the issue by converting scipy.linalg.expm to scipy.sparse.linalg.expm where appropriate.