adlbh/medinstruct-52k-arabic

---
dataset_info:
features:
- name: instruction
dtype: string
- name: input
dtype: string
- name: output
dtype: string
- name: origin_index
dtype: int64
splits:
- name: train
num_bytes: 100597846
num_examples: 52002
download_size: 46888896
dataset_size: 100597846
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
---

# Dataset Description

This dataset is a translated version of MedInstruct-52k. It contains 52k instruction-input-output triples in Arabic.
The translations were obtained using the Google Translate API.

# Tasks

The MedInstruct-52k-arabic dataset, designed for instruction fine-tuning pretrained language models on Arabic Biomedical text generation.

# Description of the Original Data

The original [MedInstruct-52k](https://github.com/XZhang97666/AlpaCare/tree/master/data) was created by [X. Zhang et al](https://arxiv.org/abs/2310.14558) following the same construction process as the Alpaca instructions dataset.

原始信息汇总

数据集概述

数据集信息

• 名称: MedInstruct-52k-arabic
• 内容: 包含52,000个指令-输入-输出三元组，内容为阿拉伯语。
• 来源: 翻译自原始的MedInstruct-52k数据集，翻译服务使用Google Translate API。

数据集特征

• 特征名称: instruction, input, output, origin_index
• 数据类型:
  • instruction: string
  • input: string
  • output: string
  • origin_index: int64

数据集划分

• 训练集:
  • 文件大小: 100,597,846字节
  • 示例数量: 52,002个

数据集大小

• 下载大小: 46,888,896字节
• 数据集总大小: 100,597,846字节

配置

• 默认配置:
  • 数据文件路径: data/train-*
  • 划分: 训练集