Llama-3.1-Nemotron-70B-Instruct 是由 NVIDIA 客製化的大型語言模型,旨在提高 LLM 生成的回應用戶查詢的實用性。

工具 70b

66.9K 4 個月前

說明

Llama-3.1-Nemotron-70B-Instruct 是由 NVIDIA 客製化的大型語言模型,旨在提高 LLM 生成的回應用戶查詢的實用性。

這個模型使用 RLHF(特別是 REINFORCE)、Llama-3.1-Nemotron-70B-RewardHelpSteer2-Preference 提示,以 Llama-3.1-70B-Instruct 模型作為初始策略進行訓練。

$ 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

Hugging Face