Do v need to provide an initial alignment for the inputs during training, in addition to the actual decoded label sequence?
2.1m questions
2.1m answers
60 comments
56.6k users