Microsoft

Beam Search Decoder Implementation

Members only

Implement beam search decoding for a language model — greedy first, then top-K beam expansion. The neural network is stubbed; you only own the decoding logic and the beam bookkeeping.