Why do we load the data once prior to the loop of epochs in the main function of training phase?

[Harry-Zhi]

Hi， Dimatura.

In your main function of file 'train.py', you first load the data once and then in the loop of every epoch you load the data again.

` loader = (data_loader(cfg, args.training_fname))

    for epoch in xrange(cfg['max_epochs']):

           loader = (data_loader(cfg, args.training_fname))`

I wonder why we first use the data_loader function before the loop (as the bold face codes show)?

Is it unnecessary or I do not get your point?
That is to say, how about just writing:
`
for epoch in xrange(cfg['max_epochs']):

           loader = (data_loader(cfg, args.training_fname))`

Could you please help me on that?