Getting the number of iterations of scipy's gmres iterative method

Question

I want to know how many iterations scipy.sparse.linalg.gmres is taking to converge, but there doesn't seem to be an argument for that. There is a maxiter argument that one could set to terminate the method, but nothing that shows the number of iterations it's taking. Can someone help me with this?

Mclin

Usually, this is not something people care about. You can use the callback argument to give a function that counts the cycles for you, though. – Uncommitted

Possible duplicate of Retrieving number of iterations that ran for sparse linear solver in SciPy – Jarman

Answer 1

To illustrate @cel's comment, you could implement a simple counter class like this:

class gmres_counter(object):
    def __init__(self, disp=True):
        self._disp = disp
        self.niter = 0
    def __call__(self, rk=None):
        self.niter += 1
        if self._disp:
            print('iter %3i\trk = %s' % (self.niter, str(rk)))

You would use it like this:

import numpy as np
from scipy.sparse.linalg import gmres

A = np.random.randn(10, 10)
b = np.random.randn(10)

counter = gmres_counter()

x, info = gmres(A, b, callback=counter)
# iter   1        rk = 0.999558746968
# iter   2        rk = 0.960490282387
# iter   3        rk = 0.945887432101
# iter   4        rk = 0.931790454216
# iter   5        rk = 0.877655067142
# iter   6        rk = 0.739596963239
# iter   7        rk = 0.677886023198
# iter   8        rk = 0.52015135005
# iter   9        rk = 0.168298366785
# iter  10        rk = 9.22692033803e-16

print(counter.niter)
# 10

Jarman