nassimjp/tsanga-alpaca-pashto-english

Tsanga Alpaca 普什图语-英语数据集是一个社区构建的、用于普什图语（Pashto）与英语之间翻译和指令跟随任务的数据集。它包含8,682个高质量的普什图语-英语句子对，以Alpaca指令格式（包含instruction、input、output字段）组织。数据集支持双向翻译（随机方向：普什图语→英语或英语→普什图语），适用于指令微调（instruction-tuning）和监督微调（SFT）。数据经过清洗、去重（基于MD5哈希去除重复项）和格式化处理，内容涵盖历史、政治、文学、诗歌、哲学、幽默、科技等多个文化丰富主题。该数据集旨在促进低资源语言（普什图语）的机器翻译、指令跟随研究以及跨语言迁移学习，可用于微调LLaMA-Factory、Axolotl等框架的模型。

Tsanga Alpaca Pashto-English Dataset is a community-built dataset for Pashto-English translation and instruction following. It contains 8,682 high-quality Pashto-English sentence pairs formatted in the Alpaca instruction format (with fields: instruction, input, output). The dataset supports bidirectional translation (randomly oriented: Pashto → English or English → Pashto) and is suitable for instruction-tuning and supervised fine-tuning (SFT). Data has been cleaned, deduplicated (using MD5 hashing), and formatted, covering culturally rich topics such as history, politics, literature, poetry, philosophy, humor, and technology. The dataset aims to facilitate machine translation, instruction-following research for low-resource languages (Pashto), and cross-lingual transfer learning, and can be used to fine-tune models with frameworks like LLaMA-Factory and Axolotl.