---
dataset_info:
  features:
  - name: source
    dtype: string
  - name: language
    dtype: string
  - name: content
    dtype: string
  - name: tokens
    dtype: int64
  - name: metadata
    struct:
    - name: type
      dtype: string
    - name: id
      dtype: int64
  splits:
  - name: train
    num_bytes: 22370299
    num_examples: 20483
  download_size: 3415442
  dataset_size: 22370299
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: mit
language:
- en
tags:
- code
size_categories:
- 1M<n<10M
---
🤖 MUD-Code3 – 2M Real‑World Code Tokens!
This dataset contains 20483 code snippets (mixture of real and high-quality synthetic) totaling 2,000,018 tokens.

Languages: Python (and some others)
Sources: Real open‑source code (when available) plus enhanced synthetic
Use cases: Fine‑tuning code models, program synthesis, code search
📖 How to Use
from datasets import load_dataset
dataset = load_dataset("CompiwerAI/MUD-Code3")
print(dataset["train"][0])

📊 Stats
Metric	Value
Total Documents	20483
Total Tokens	2,000,018
File Size (raw)	~23 MB
📜 License & Citation
MIT License.
If you use this dataset, please cite:

@misc{mud-code3-2026,
  author = {CompiwerAI},
  title = {MUD‑Code3: A 2M‑token Real‑Code Dataset},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/CompiwerAI/MUD-Code3}
}