amalia-llm/PT-Culture_Data
收藏资源简介:
葡萄牙文化监督微调数据集是一个用于欧洲葡萄牙语(PT-PT)文化知识的监督微调数据集,旨在教导模型葡萄牙文化的传统、人物、地点和表达方式。该数据集包含两个版本,共计216,832个示例,以对话式监督微调格式组织,涵盖十个文化领域:人物、视听、遗产、体育、地理、文学、美食、谚语、节日和假日。数据涵盖七种任务类型:验证、短问答和长问答、文化背景推理、实体比较、多轮对话(角色多样化)以及谚语含义解释。提供两个文件:culture_data_sft.jsonl(v1)和culture_data_sft_v2.jsonl(v2)。如需使用全部数据,建议合并两个文件。该数据集是AMALIA项目的一部分,用于后训练AMALIA模型。
A supervised fine-tuning dataset for European Portuguese (PT-PT) cultural knowledge, built to teach models the traditions, figures, places, and expressions of Portuguese culture. The dataset contains 216,832 examples across two versions in conversational SFT format, organised into ten cultural domains: Personalities, Audiovisual, Heritage, Sports, Geography, Literature, Gastronomy, Proverbs, Festivals, and Holidays. The data spans seven task types: Verification, short and long question-answering, cultural-context reasoning, entity comparison, multi-turn dialogue (persona-diversified) and proverb meaning explanation. Two files are available: culture_data_sft.jsonl (v1) and culture_data_sft_v2.jsonl (v2). For training on all data, use both files combined. This dataset is provided as part of the AMALIA project and is included in the data mix used to post-train the AMALIA model.




