Allow to set the top_p parameter to a higher value, which will allow the model to consider a larger number of possible completions before selecting the final output.

Also allow to to set the repetition_penalty parameter to a lower value, which will decrease the model's tendency to repeat phrases and words.