python - Memory leak using scipy.eigs -
i have set of large sparse matrices i'm trying find eigenvector corresponding eigenvalue 0 of. structure of underlying problem know solution must exist , 0 largest eigenvalue.
to solve problem use scipy.sparse.linalg.eigs
arguments:
val, rho = eigs(l, k=1, = 'lm', v0=init, sigma=-0.001)
where l
matrix. call multiple times different matrices. @ first having severe problem memory usage increases call function more , more times. seems eigs
doesn't free memory should. solved calling gc.collect()
each time use eigs
.
but worry internally memory isn't being freed, naively expect using arnoldi shouldn't use more memory algorithm progresses, should storing matrix , current set of lanczos vectors, find memory usage increases while code still inside eigs
function.
any ideas?
Comments
Post a Comment