DistributedHelper CI is failing

[AntonioCarta]

@lrzpellegrini check this error.

[lrzpellegrini]

Hi Antonio, it seems a spurious error. Alas, when shutting down a process that used the PyTorch distributed training api + unittest, the process closing procedures may raise an error. This is not an error of the distributed training helper.

One easy way to circumvent this it to re-run the unit tests that raised an error again. I think I can adapt https://github.com/ContinualAI/avalanche/blob/master/tests/run_dist_tests.py to achieve this.

[AntonioCarta]

Maybe we should run those tests without unittest?