The nnlf method (Negative log-likelihood function) exists to do exactly this:

import numpy as np
from scipy.stats import norm

data = [1,2,3,4,5]
m,s = norm.fit(data)
log_likelihood = -norm.nnlf([m,s], data)