Llama-3.1-Nemotron-70B-Instruct 是一個由 NVIDIA 客製化的大型語言模型,旨在提升 LLM 生成的回應對於使用者查詢的實用性。

工具 70b

60.7K 3 個月前

說明文件

Llama-3.1-Nemotron-70B-Instruct 是一個由 NVIDIA 客製化的大型語言模型,旨在提升 LLM 生成的回應對於使用者查詢的實用性。

這個模型使用 RLHF (具體來說是 REINFORCE) 進行訓練,以 Llama-3.1-Nemotron-70B-RewardHelpSteer2-Preference prompts 提示,並以 Llama-3.1-70B-Instruct 模型作為初始策略。

$ ollama run nemotron
>>> How many r in strawberry?
A sweet question!

Let's count the "R"s in "strawberry":

1. S
2. T
3. R
4. A
5. W
6. B
7. E
8. R
9. R
10. Y

There are **3** "R"s in the word "strawberry".

參考資料

Hugging Face