Beam Search Decoding with Joint Probabilities

microsoft

Given a layered state-transition graph for an autoregressive decoder, implement both greedy decoding and beam-search decoding.

The initial state is 0. Starting from state 0, generate exactly D ...

Example

Unlock to view complete problem details

and practice with sample input/output

Was this article helpful?

View Test Cases & Run Code requires membership

Standard Input
Execution Result: