Found that train.num_workers hyperparameter from config not used in train dataloader initialization.
It can cause errors while using Windows OS (As example, OSError: [WinError 6]), which not correctly supporting parallelization in pytorch, so it needs to num_workers=0 used.
Found that
train.num_workers
hyperparameter from config not used in traindataloader
initialization.It can cause errors while using Windows OS (As example,
OSError: [WinError 6]
), which not correctly supporting parallelization in pytorch, so it needs tonum_workers=0
used.