Ollama has a number of parameters that help in generation of coherent text. As I understand, these parameters are for fine tuning text output irrespective of which ONE of the LLM’s is picked up from the ollama library.

Parameters such as mirostat sampling, top_p, top_k , tfs_z and repeat_penalty have the same objective ie removing confused and repetitive text output and generate a coherent text. As, I also understand, these parameters should be considered as mutually exclusive in the sense that if I am trying mirostat sampling then do not try top_p or top_k or tfs_z. And if trying top_p, do not try the rest.

I have, so to say, two related questions: Is my assumption about mutual exclusivity correct? And if so, how should I begin to try to set which of the parameters first and then which should be the second and so on. OR, are there any rough guidelines as to which one of these parameters would be most influential and which one the least?