Returns the log probabilities of the top candidate tokens at each generation step. The model's chosen token might not be the same as the top candidate token at each step. Specify the number of candidates to return by using an integer value in the range of 1-20.
logprobs: number