In which function it is best way to use the temperature parameter .from_pretrained() or .generate()

🤗Transformers
1

in which function is the best way to use the temperature parameter .from_pretrained or .generate()
It seems that in both functions we can pass the temperature parameter. Which is the best way to do this?

Model = AutoModelForCausalLM.from_pretrained(model_name,
                                                                   device_map="auto", 
                                                                   temperature=0.1,
                                                                   do_sample=True,
                                                                   torch_dtype=torch.bfloat16)

or

Model.generate(**model_input,max_length=1000, temperature=0.1,do_sample=True,)