About the contextual dialogue ability (or the memory ability)

[mcddddz]

Thank you for the llama3 model. Now I'm trying some inference on Meta-Llama-3.1-8B, but it seems that it don't have the contextual dialogue ability. To be more specific, when I talked to it at second time, it have already forgotten the content when I first talked to it, althrough I just talked to it just a few seconds ago. Is there any wrong in my operation, or just because Meta-Llama-3.1-8B don't have a memory ability? Maybe adding the context I've talked to it to the new conversation can improve it, but is it the correct way to use this language model?