LumiOpen/poro2-instruction-collection
收藏Hugging Face2025-06-19 更新2025-07-05 收录
下载链接:
https://hf-mirror.com/datasets/LumiOpen/poro2-instruction-collection
下载链接
链接失效反馈官方服务:
资源简介:
Poro2 SFT Mix是一个包含英文和芬兰语的大规模合成指令数据集,适用于SFT训练。数据集主要由翻译自Tulu3 SFT Mixture的提示组成,并补充了来自Open Assistant 2和Avoin Avustaja的顶级芬兰语对话。英文数据通过Llama-3.3-70B-Instruct模型生成,并包含了多轮对话。此外,数据集还包含了来自EuroParl数据集的英芬互译样本。
Poro2 SFT Mix is a large collection of mostly synthetic instruction data in English and Finnish, suitable for SFT training. The dataset consists mainly of prompts translated from the Tulu3 SFT Mixture, supplemented with top Finnish conversations from Open Assistant 2 and Avoin Avustaja. The English data is generated using the Llama-3.3-70B-Instruct model, including multi-turn conversations. Additionally, the dataset includes English-Finnish translation samples from the EuroParl dataset.
提供机构:
LumiOpen



