Do v need to provide an initial alignment for the inputs during training, in addition to the actual decoded label sequence?
1.4m articles
1.4m replys
5 comments
57.0k users