bogus_script.py

import numpy as np
rs = np.random.RandomState(seed=5566)
n_conditions = 10

# Here we simulate a complex computation, for example, analogy of the magnitude
# of gradient decent which expects to be strictly positive. But from the result
# we find that it seems to be sometimes negative, we wish to find out when and
# what condition our program produces bogus ouput.
#
# This is the case to use pdb and condition break point
result_sum = 0.0

# To enable pdb debug, uncomment the following two lines:
#
# import pdb
# pdb.Pdb(skip=['numpy.*']).set_trace()
#
# setting skip=['some modules'] so we could use 's' (step) into our own code
# instead of debugging every module we use.
for i in range(n_conditions):
    complex_out = rs.normal(loc=3.0, scale=0.8, size=1000)
    # we found the result is not right sometimes
    result_sum += np.sum(complex_out)