Small issue in nucleus sampling explanation + implementation

[ascher8]

In the file <labml_nn/sampling/nucleus.py> there seem to be two issues.
"That is, we pick the highest probable tokens until the sum of their probabilities is less that $p$." should be more than
And in the code implementation, line 62 should be flipped to a >= from what it is currently:
nucleus = cum_sum_probs < self.p

I'm new to github and just learned what nucleus sampling is 10 minutes ago, so sorry if this Issue was improperly formatted or wrong or anything.