Hi,
I have a question about step used in Transformer Decoder.
in PositionalEncoding, it uses ‘step’ under a certain condition.
I read already comments for this
step (int or NoneType): If stepwise (``seq_len = 1``), use
the encoding for this position.
but still I’m wonder in which case we are using it.
so can someone explain it for give some links of paper/blog to related it?
BTW, I can’t find any code not to set it as ‘None’, is there somewhere else?
Thanks,