Closed orsharir closed 7 years ago
Actually the current padding argument could be 'SAME', 'VALID' or a list with integer values to specify the actual paddings, is this okay @orsharir ? (it's orthogonal of course to the upper/lower case issue)
Yes, that's the correct behavior we've talked about. The upper/lower case issue is separate from that. If it's now fixed, you can close this issue.
Notice that in Tensorflow the padding api is specified in uppercase, i.e.
'SAME'
and'VALID'
, while in Keras it is in lowercase, i.e.'same'
and'valid'
. Our API should adhere to this convention. You could probably use this Keras's method to handle this change (or simply write your own).