Explanation of the Options
The responses are composed out of tokens, which are common sequences of characters. A token is often one word, but it might be a part of a word or a punctuation mark. The maximum tokens are shared between the prompt and the response.
The temperature determines how tokens are sampled. A higher temperature will return more creative results, while a lower value will return a more definite answer.
The value of top_p determines how likely a given token must be in order to be considered. A top_p of 0.5 means that only the 50% most likely tokens are considered, while a value of 0.1 means that the 10% most likely are considered
The frequency penalty determines how likely tokens are to appear in a sequence in which they already appear in the text. A low value makes verbatim repetition less likely.
The presence penalty determines how likely new tokens are to appear in the response. A low value makes new words more likely. A high value makes new words less likely.