Repeat each value according to num hyps. Args: source_enc_len: source encoder length; int [batch]. num_hyps_per_beam: number of hypotheses Returns: New version of source_enc_len; int [batch * num_hyps_per_beam]. Target_batch is (num_hyps_per_beam * batch). Exa
(self, source_enc_len, num_hyps_per_beam)
source not stored for this graph (policy: none)
no outgoing calls