cvx.simulator.backtest_many with parallel=True get stuck on Linux

[andersontigerzhang]

I run cvxportfolio (1.3.0), SinglePeriodOptimization on Linux with a few policies. It runs OK when the policy is simple policies such as AdaptiveRebalance, but it gets stuck when I change to SinglePeriodOptimization. I googled the answer, it looks like it is the behavior on multiprocessing.Pool.starmap on simulator.py, line 707. If I changed the code from:
with Pool(initializer=_mp_init, initargs=(Lock(),)) as p:
to
with get_context('spawn').Pool(initializer=_mp_init, initargs=(Lock(),)) as p:
then it all worked.

I could not find a solution without changing the library code. I tried to put multiprocess.set_start_method('spawn') as the first statement right after:
if name == 'main':

as suggested by google search, the process still stuck. Anyone have any idea how to solve this problem? I understand it is not quite a cvxportfolio issue.