deep learning - is loss value should be divided by `seq_length`

Question

I\'m new to deep learning and I\'m looking at train.py here: https://github.com/huanghao-code/VisRNN_ICLR_2016_Text/blob/master/train.py.

I run this code and the loss val