As sub-processes fork the random generator from their parent process, it might happen that different sub processes obtain the same sequence of random numbers in different processes. Depending on the backend it might be an issue. In general it would be advantageous to make it explicit, e.g.
def set_time_pid_seed():
t = int(os.getpid() + time.time())
np.random.seed(t)
random.seed(t)
As sub-processes fork the random generator from their parent process, it might happen that different sub processes obtain the same sequence of random numbers in different processes. Depending on the backend it might be an issue. In general it would be advantageous to make it explicit, e.g.
or even better, to pass different seeds from the parent process to the forked child processes. The latter would give us reproducability, see also (https://joblib.readthedocs.io/en/latest/auto_examples/parallel_random_state.html)