---
tags:
- capybara
- subset
- conversational
- sft
license: apache-2.0
size_categories:
- n<1K
---

# Capybara-100

A 100-sample subset of the [LDJnr/Capybara](https://huggingface.co/datasets/LDJnr/Capybara) dataset.

## Description

This is a small, reproducible subset of the Capybara conversational dataset containing 100 randomly selected examples. It was created by shuffling the original dataset with seed `42` and selecting the first 100 examples.

## Dataset Details

- **Source**: [LDJnr/Capybara](https://huggingface.co/datasets/LDJnr/Capybara)
- **Total original examples**: 16,006
- **Subset size**: 100
- **Shuffle seed**: 42
- **Columns**: `source`, `conversation`

## Usage

```python
from datasets import load_dataset

ds = load_dataset("lewtun/capybara-100", split="train")
print(len(ds))  # 100
```

## Format

The dataset is stored as a JSONL file with the following schema:
- `source`: String indicating the source category (e.g., "General-Instruct", "Dove")
- `conversation`: List of turns, each with `input` and `output` fields