Pyper

Pyper is a framework for generating high-quality synthetic data for LLM instruction tuning. It combines several research approaches to create diverse and high-quality training datasets.

Features

• General Seed Generation: Generate seed data across various academic disciplines
• Knowledge-Driven Generation: Create datasets based on specific domain knowledge by supplying ground truth information
• Data Fission: Expand small seed datasets through iterative generation using LLM

Installation

Pyper uses Poetry for dependency management. To install:

# Install poetry
curl -sSL https://install.python-poetry.org | python3 -

# Clone the repository
git clone https://github.com/yourusername/pyper.git
cd pyper

# Install dependencies
poetry install

Usage

Environment Setup

First, set your OpenAI API key:

export OPENAI_API_KEY=your-openai-api-key

General Seed Generation

Generate seed data for a specific academic discipline:

python run.py generate --mode general --discipline mathematics --num-tasks 50

Example disciplines include mathematics, physics, chemistry, etc. Use --help to see all options.

Knowledge-Driven Generation

Generate data using specific domain knowledge:

python run.py generate --mode knowledge \
    --knowledge-path ./path/to/knowledge.txt \
    --num-tasks 50

The knowledge file should contain the ground truth information in plain text format.

Data Fission

Expand an existing seed dataset:

python run.py fission \
    --num-tasks 50 \
    ---seed-path path/to/seed \

To see all available options, use --help flag

Research & Citations

Pyper's implementation is based on several key research papers:

@misc{selfinstruct,
  title={Self-Instruct: Aligning Language Model with Self Generated Instructions},
  author={Wang, Yizhong and Kordi, Yeganeh and Mishra, Swaroop and Liu, Alisa and Smith, Noah A. and Khashabi, Daniel and Hajishirzi, Hannaneh},
  journal={arXiv preprint arXiv:2212.10560},
  year={2022}
}

@misc{alpaca,
  author = {Rohan Taori and Ishaan Gulrajani and Tianyi Zhang and Yann Dubois and Xuechen Li and Carlos Guestrin and Percy Liang and Tatsunori B. Hashimoto },
  title = {Stanford Alpaca: An Instruction-following LLaMA model},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/tatsu-lab/stanford_alpaca}},
}

@article{luo2023wizardmath,
  title={WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct},
  author={Luo, Haipeng and Sun, Qingfeng and Xu, Can and Zhao, Pu and Lou, Jianguang and Tao, Chongyang and Geng, Xiubo and Lin, Qingwei and Chen, Shifeng and Zhang, Dongmei},
  journal={arXiv preprint arXiv:2308.09583},
  year={2023}
}

Additional research:

• Generalized Instruction Tuning (GLAN)