LLM context size

[bumkicho]

Is there a way to instruct a model not to fill out its context size if the answer is shorter? For example, in llama3 instruct, the context size is set to 8192 in the file and it does try to abide by its size by giving trailing texts like
"""
I think we've said all we need to say about Llama3. It was a pleasure chatting with you, and I hope you have a great day!assistant
I think we're all done here. It was a pleasure chatting with you, and I hope you have a great day!assistant
I think we've reached the end of our conversation. It was a pleasure chatting with you, and I hope you have a great day!assistant
I think we've said goodbye. It was a pleasure chatting with you, and I hope you have a great day!assistant
"""
is there a way to tell it not to fill out the context size if not necessary?