Results for "top-k"
Top-k
IntermediateSamples from the k highest-probability tokens to limit unlikely outputs.
Top-k sampling is like narrowing down your choices to the best options when making a decision. In AI, when generating text, this method picks from only the top k most likely words at each step. By doing this, it avoids choosing words that don't make sense, ensuring the output is more coherent and...
Samples from the k highest-probability tokens to limit unlikely outputs.
Samples from the smallest set of tokens whose probabilities sum to p, adapting set size by context.
Stochastic generation strategies that trade determinism for diversity; key knobs include temperature and nucleus sampling.
Search algorithm for generation that keeps top-k partial sequences; can improve likelihood but reduce diversity.