jescy525/aether-sft-v2-conversation

这是一个AETHER家族的监督微调（SFT）数据集，属于conversation组，专门用于训练对话生成模型。数据集采用JSONL格式存储ChatML消息，包含消息列表（角色包括system、user、assistant和对应内容）、任务类型（如function_calling、code、reasoning_cot等）、来源数据集ID、语言（如英语、法语）和系统来源信息。数据经过MinHash去重处理（阈值为0.85），以确保数据质量。该数据集由prepare_sft.py流水线生成（日期2026-05-25），旨在支持多语言对话任务，适用于文本生成和对话系统开发。This is a supervised fine-tuning (SFT) dataset from the AETHER family, belonging to the conversation group, designed for training conversational generation models. The dataset is stored in JSONL format with ChatML messages, including message lists (with roles such as system, user, assistant and corresponding content), task types (e.g., function_calling, code, reasoning_cot, etc.), source dataset IDs, languages (e.g., English, French), and system source information. The data has undergone MinHash deduplication (threshold 0.85) to ensure quality. Generated by the prepare_sft.py pipeline (date 2026-05-25), it supports multilingual conversation tasks and is suitable for text generation and dialogue system development.