instruct_pashto
收藏资源简介:
Instruct Pashto 数据集是一个专门为普什图语指令遵循任务设计的对话数据集。该数据集包含多个知名指令对话数据集(如Wikipedia、Databricks的Dolly和MosaicML Instruct-v3)的普什图语翻译内容,这些原始数据均基于CC BY-SA 3.0许可证发布。数据集规模小于1万样本,适用于文本生成任务,特别关注医疗、指令和聊天等应用场景。数据采用标准的jsonl格式存储,每条记录包含一个消息列表,其中明确区分用户角色和助理角色的对话内容,格式化为适合微调对话式大语言模型(LLMs)的结构。数据集的主要目的是支持在普什图语环境下开发能够理解和遵循指令的对话AI模型,填补了低资源语言在指令微调数据方面的空白。所有翻译和重新格式化的普什图语内容由Nsibullah Nassim于2026年整理并发布。
The Instruct Pashto dataset is a conversational dataset specifically designed for Pashto language instruction following tasks. It contains Pashto translations of several well-known instruction conversational datasets, including Wikipedia, Databricks' Dolly, and MosaicML Instruct-v3. All original datasets are released under the CC BY-SA 3.0 license. The dataset has fewer than 10,000 samples, is suitable for text generation tasks, and focuses specifically on application scenarios such as healthcare, instruction following, and chat. The data is stored in the standard JSONL format. Each record includes a list of messages that clearly distinguish between user and assistant roles, formatted to be compatible with fine-tuning conversational large language models (LLMs). The primary purpose of this dataset is to support the development of conversational AI models that can understand and follow instructions in Pashto environments, filling the gap in instruction fine-tuning data for low-resource languages. All translated and reformatted Pashto content was compiled and released by Nsibullah Nassim in 2026.
数据集卡片:Instruct Pashto
- 语言:普什图语(ps)
- 任务类别:文本生成
- 标签:医学、指令、聊天
- 数据集规模:少于10,000条样本(n<10K)
数据集描述
该数据集包含指令式对话数据集的普什图语翻译版本,专门用于微调面向普什图语指令遵循任务的对话式大型语言模型(LLMs)。数据以标准 jsonl 格式存储,每条记录包含一个 messages 列表,列表中的消息依次标注为 user 和 assistant 角色。
数据格式示例
json {"messages": [ {"role": "user", "content": "دې پوښتنې ته ځواب ورکړئ: ..."}, {"role": "assistant", "content": "ځواب 50 دی."} ]}
许可与归属
该数据集源自原始采用 CC BY-SA 3.0 许可的指令数据集。原始来源包括:
- Wikipedia — CC BY-SA 3.0
- Dolly (Databricks) — CC BY-SA 3.0
- MosaicML Instruct-v3 — CC BY-SA 3.0
所有翻译并重新格式化的普什图语内容 © 2026 Nsibullah Nassim。
引用
如使用该数据集,请引用:
@misc{instruct_pashto, author = {Nsibullah Nassim}, title = {Instruct Pashto Dataset}, year = {2026}, publisher = {HuggingFace Datasets}, howpublished = {https://huggingface.co/datasets/nassimjp/instruct_pashto}, }
基于:
@misc{mosaicml2023instruct-v3, author = {MosaicML}, title = {MosaicML Instruct-v3 Dataset}, year = {2023}, publisher = {HuggingFace Datasets}, howpublished = {https://huggingface.co/datasets/mosaicml/instruct-v3}, }




