You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Minor corrections in documentation for Nucleus Sampling implementation #293
Hello! Thank you for sharing the clean and well-documented implementation of nucleus sampling. While reading through the explanation, I noticed two small errors in the written description that could be slightly confusing for readers.
In the section that begins:
"That is, we pick the highest probable tokens until the sum of their probabilities is less that $p$."
There are two issues:
Logical oversight: The condition should be "is not less than $p$" (or equivalently, "is at least $p$"), since we want the smallest set of tokens whose cumulative probability reaches or exceeds $p$.
Activity