Dmitry Burdeiny

@darissa Please discard all changes and checkout the commit a53b52d. Then open the script `train.py`, find all occurrences of `num_workers=4` and replace them by `num_workers=0` (there are 3 occurrences). Launch...

OK. After initializing `num_workers=0`, the script should work on Windows, too (I hope). (The issue was likely related to this one: https://discuss.pytorch.org/t/dataloader-multiprocessing-error-cant-pickle-odict-keys-objects-when-num-workers-0/43951)

What is the hardware you are trying to train on? (When you tried on Windows)

Yes, you can do that by wrapping the model into `nn.DataParallel`. Check out the parallelism tutorials: [1](https://pytorch.org/tutorials/beginner/former_torchies/parallelism_tutorial.html), [2](https://pytorch.org/tutorials/beginner/blitz/data_parallel_tutorial.html). (Currently the code in this repo does not use `nn.DataParallel`.)