Llama-Nemotron-Post-Training-Dataset-v1
收藏资源简介:
Llama-Nemotron后训练数据集v1是一个包含支持数学、代码、泛推理和指令跟随能力提升的数据集,由NVIDIA发布,用于提升Llama-3.3-Nemotron-Super-49B-v1和Llama-3.1-Nemotron-Nano-8B-v1模型的表现。数据集分为SFT和RL两种配置,并包含了数学、代码、科学、聊天和安全等多个类别的数据文件。数据集遵循CC-BY-4.0许可,部分数据遵循ODC-BY和CC-BY-SA许可。
The Llama-Nemotron Post-Training Dataset v1, developed and released by NVIDIA, is a curated dataset designed to enhance mathematical, coding, general reasoning, and instruction-following capabilities, with the primary objective of improving the performance of Llama-3.3-Nemotron-Super-49B-v1 and Llama-3.1-Nemotron-Nano-8B-v1 models. The dataset is structured into two configurations: SFT and RL, and includes data files across multiple categories such as mathematics, coding, science, chat, and safety. This dataset is licensed under CC-BY-4.0, with portions of its data licensed under ODC-BY and CC-BY-SA respectively.




