sow-mini-dataset-en-vi
收藏资源简介:
sow-mini-dataset 是 minimind 数据集翻译为英语和越南语的版本。数据集包含两种格式:一种保留原始 minimind schema 的翻译记录,位于 <lang>/<file>/shard-NNNNN.jsonl;另一种是 QA 格式,位于 rwkv/<lang>/<file>/shard-NNNNN.jsonl,每条记录包含一个 text 字段,可直接用于 token 化处理。每个样本还包含 tools(函数调用 schema)和 gt(精确匹配答案)字段,这些字段被原样保留,未翻译。数据集按语言和文件分片存储,每个分片为 JSONL 格式。根据翻译过程的统计信息,英语和越南语各有 2 个分片。该数据集适用于多语言翻译、指令微调、函数调用理解等任务。
The sow-mini-dataset is a translated version of the minimind dataset into English and Vietnamese. It contains two formats: one preserving the original minimind schema with translated records stored at <lang>/<file>/shard-NNNNN.jsonl; the other is a QA format at rwkv/<lang>/<file>/shard-NNNNN.jsonl, where each record has a text field ready for tokenization. Each sample also includes tools (function call schema) and gt (exact match answer) fields, which are kept untranslated. The dataset is sharded by language and file, each shard in JSONL format. According to translation statistics, English and Vietnamese each have 2 shards. This dataset is suitable for tasks such as multilingual translation, instruction fine-tuning, and function call understanding.




