Stochastic Decoding

[leonlenk]

• This will be demoed on Thursday so use the attention transformer as a baseline
• show the differences between and how to use a greedy algorithm, k-top, p-top, temperature, and beam search to sample from the model's output
• Have the members impliment each algorithm themselves