RobCzikkel/DoctorGPT

---
language:
- en
size_categories:
- 10K<n<100K
task_categories:
- conversational
pretty_name: Doctor & Patient
dataset_info:
features:
- name: prompt
dtype: string
- name: input_ids
sequence: int32
- name: length
dtype: int64
- name: attention_mask
sequence: int8
splits:
- name: train
num_bytes: 42127351.778204426
num_examples: 13125
- name: test
num_bytes: 10534245.221795576
num_examples: 3282
download_size: 10917910
dataset_size: 52661597.0
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
tags:
- biology
- medical
---

### Dataset
This is an edited and tokenized version of the MedQuad-MedicalQnADataset dataset by keivalya.
The original dataset contains 16K+ questions and answers between patient and doctor, which have been converted into a full prompt to train BioGPT by Microsoft.

##### Tokenizer used
microsoft/BioGPT-Large (BPE tokenizer)

### Full prompt

```py
prompt = f"""You are a helpful AI Doctor who answers medical questions. Below is a question from a patient. Your task is to answer the questions as truthfully as you can.

### Patient:
{sample['Question']}

### Doctor:
{sample['Answer']}"""
```

### Notes
Since bioGPT has a max input of 1024, the full prompt was truncated to stay below this limit.
The truncation strategy I used made sure that only full sentences were produced.

Please note that this dataset is for research/testing only, it should not be used in a real setting or used to give medical advice to people.

原始信息汇总

数据集概述

基本信息

• 语言: 英语
• 大小类别: 10K<n<100K
• 任务类别: 对话
• 美观名称: Doctor & Patient

数据集详情

• 特征:
  • prompt: 字符串类型
  • input_ids: 序列类型，int32
  • length: int64类型
  • attention_mask: 序列类型，int8
• 分割:
  • train: 42127351.778204426字节，13125个样本
  • test: 10534245.221795576字节，3282个样本
• 下载大小: 10917910字节
• 数据集大小: 52661597.0字节

配置

• 默认配置:
  • 数据文件:
    • train: data/train-*
    • test: data/test-*

标签

• biology
• medical

数据集描述

• 该数据集是MedQuad-MedicalQnADataset的编辑和标记化版本，由keivalya提供。
• 原始数据集包含16K+的医患问答，已转换为完整提示以训练Microsoft的BioGPT。

使用的标记器

• microsoft/BioGPT-Large (BPE tokenizer)

完整提示

python prompt = f"""You are a helpful AI Doctor who answers medical questions. Below is a question from a patient. Your task is to answer the questions as truthfully as you can.

Patient:

{sample[Question]}

Doctor:

{sample[Answer]}"""

注意事项

• 由于BioGPT的最大输入为1024，完整提示已被截断以保持在限制以下。
• 截断策略确保只生成完整句子。
• 该数据集仅用于研究/测试，不应在实际环境中使用或用于向人们提供医疗建议。