---
license: mit
language:
- en
tags:
- debug
- chat
- conversational
- ultrachat
size_categories:
- n<10
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
dataset_info:
  features:
  - name: prompt
    dtype: string
  - name: prompt_id
    dtype: string
  - name: messages
    list:
    - name: content
      dtype: string
    - name: role
      dtype: string
  splits:
  - name: train
    num_bytes: 81146
    num_examples: 10
  - name: validation
    num_bytes: 40969
    num_examples: 8
  download_size: 77089
  dataset_size: 122115
---

# femto-ultrachat

A tiny 10-sample subset of [UltraChat 200k](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k) designed specifically for debugging purposes.

## Purpose

This dataset contains only the first 10 samples from the UltraChat 200k dataset (train_sft split), making it ideal for:
- Quick debugging of chat/conversational AI pipelines
- Testing chat model fine-tuning code without downloading large datasets
- Rapid prototyping of chat interfaces
- CI/CD testing for conversational AI systems
- Validating chat template implementations

## Dataset Structure

The dataset has the same structure as the original UltraChat dataset:
- `messages`: List of message dictionaries, each containing:
  - `role`: Either "user" or "assistant"
  - `content`: The message text
- `prompt_id`: Unique identifier for the conversation

### Example

```python
{
    "messages": [
        {"role": "user", "content": "What is the capital of France?"},
        {"role": "assistant", "content": "The capital of France is Paris."},
        ...
    ],
    "prompt_id": "..."
}
```

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("Butanium/femto-ultrachat", split="train")
print(f"Dataset size: {len(dataset)} samples")

# Access first conversation
conversation = dataset[0]["messages"]
for message in conversation:
    print(f"{message['role']}: {message['content']}")
```

### Testing Chat Templates

This dataset is perfect for testing chat templates:

```python
from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-chat-hf")
dataset = load_dataset("Butanium/femto-ultrachat", split="train")

# Test apply_chat_template
formatted = tokenizer.apply_chat_template(
    dataset[0]["messages"],
    tokenize=False
)
print(formatted)
```

## Source

This dataset is derived from [UltraChat 200k](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k) and inherits its MIT license.

## Citation

If you use UltraChat in your research, please cite:

```bibtex
@misc{ding2023enhancing,
      title={Enhancing Chat Language Models by Scaling High-quality Instructional Conversations}, 
      author={Ning Ding and Yulin Chen and Bokai Xu and Yujia Qin and Zhi Zheng and Shengding Hu and Zhiyuan Liu and Maosong Sun and Bowen Zhou},
      year={2023},
      eprint={2305.14233},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}
```