---
license: cc-by-4.0
configs:
- config_name: v1.0
  data_files:
  - split: train
    path: v1.0/train-*
  - split: dev
    path: v1.0/dev-*
  - split: test
    path: v1.0/test-*
dataset_info:
  config_name: v1.0
  features:
  - name: ID
    dtype: string
  - name: text
    dtype: string
  - name: output
    dtype: string
  - name: meta
    struct:
    - name: alert-type
      sequence: string
    - name: domain
      sequence: string
    - name: output-reference
      sequence: string
    - name: output-type
      sequence: string
    - name: perspective
      sequence: string
    - name: question-reference
      sequence: string
    - name: source-to-answer
      sequence: string
    - name: task
      sequence: string
    - name: time-dependency
      dtype: bool
  splits:
  - name: train
    num_bytes: 559676
    num_examples: 300
  - name: dev
    num_bytes: 565546
    num_examples: 300
  - name: test
    num_bytes: 759013
    num_examples: 400
  download_size: 2084518
  dataset_size: 1884235
task_categories:
- text-generation
language:
- ja
size_categories:
- n<1K
---

## 概要

llm-jp-instructionsは人手により作成されたインストラクションデータセットです。

## Overview

llm-jp-instructions is a manually created instruction dataset.

## Usage

```
from datasets import load_dataset

# load train, dev and test splits of v1.0
v1_train = load_dataset("llm-jp/llm-jp-instructions", data_dir="v1.0", split="train")
v1_dev = load_dataset("llm-jp/llm-jp-instructions", data_dir="v1.0", split="dev")
v1_test = load_dataset("llm-jp/llm-jp-instructions", data_dir="v1.0", split="test")
```