Bug in tut_intro.ipynb when calling cv.MultiSim

[TristanFauvel]

Describe the bug

In the last cell of the tutorial tut_intro.ipynb , when running msim = cv.MultiSim([s1, s2]), I get the following error message:

BrokenProcessPool: A process in the process pool was terminated abruptly while the future was running or pending.

The full output is:

---------------------------------------------------------------------------
BrokenProcessPool                         Traceback (most recent call last)
Cell In[17], line 22
     20 s2 = cv.Sim(pars, interventions=protect_elderly, label='Protect the elderly')
     21 msim = cv.MultiSim([s1, s2])
---> 22 msim.run()
     23 fig = msim.plot(['cum_deaths', 'cum_infections'])

File c:\users\tfauvel\documents\covasim\covasim\run.py:172, in MultiSim.run(self, reduce, combine, **kwargs)
    170 # Run
    171 kwargs = sc.mergedicts(self.run_args, kwargs)
--> 172 self.sims = multi_run(sims, **kwargs)
    174 # Reduce or combine
    175 if reduce:

File c:\users\tfauvel\documents\covasim\covasim\run.py:1496, in multi_run(sim, n_runs, reseed, noise, noisepar, iterpars, combine, keep_people, run_args, sim_args, par_args, do_run, parallel, n_cpus, verbose, retry, **kwargs)
   1494         raise RuntimeError(errormsg) from E
   1495     else: # For all other runtime errors, raise the original exception
-> 1496         raise E
   1497 except pkl.PicklingError as E:
   1498     parallelizer = par_args.get('parallelizer')

File c:\users\tfauvel\documents\covasim\covasim\run.py:1477, in multi_run(sim, n_runs, reseed, noise, noisepar, iterpars, combine, keep_people, run_args, sim_args, par_args, do_run, parallel, n_cpus, verbose, retry, **kwargs)
   1475 kw = dict(iterkwargs=iterkwargs, kwargs=kwargs, **par_args)
   1476 try:
...
    392     finally:
    393         # Break a reference cycle with the exception in self._exception
    394         self = None

BrokenProcessPool: A process in the process pool was terminated abruptly while the future was running or pending.

Note that the following works fine:

s1 = cv.Sim(pars, label='Default')
s2 = cv.Sim(pars, interventions=protect_elderly, label='Protect the elderly')

Platform (please complete the following information):

• Covasim version: 3.1.4
• OS: Windows 10
• Other platform information: using Anaconda Python

[cliffckerr]

Hi @TristanFauvel , sorry you're encountering an issue! In general, on Windows, any calls to multiprocess need to happen within the __main__ block; see the answer here:
https://docs.idmod.org/projects/covasim/en/latest/faq.html#why-do-parallel-simulations-fail-on-windows-or-in-jupyter-notebooks

However, this doesn't usually happen from within a Jupyter notebook, so I'm surprised you encountered that here. While adding the __main__ block should fix it, it's possible that upgrading your Jupyter/IPython kernel might as well. Let us know if that works, and if not then we can investigate further and potentially update the FAQ.

[TristanFauvel]

Hi @cliffckerr,
thanks for the quick reply, and for the great package. Actually when running the example in the FAQ I don't get an error, even when not using a __main__ block. By contrast, adding a __main__ block does not solve the issue in tut_intro.ipynb.