Model Card for Phi-4-Argunaut-1-SPIN

This model is a fine-tuned version of DebateLabKIT/Phi-4-Argunaut-1-SFT. It has been trained using TRL and vLLM. Checkpoints are tagged.

📘 HF Blog Article

Quick start

from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="DebateLabKIT/Llama-3.1-Argunaut-1-8B-SPIN", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

This model was trained with Self-Play Fine-Tuning (SPIN), a method introduced in Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models.

More details about the training procedure can be found in the blog post.

Framework versions

• TRL: 0.14.0
• Transformers: 4.46.3
• Pytorch: 2.4.1
• Datasets: 3.1.0
• Tokenizers: 0.20.3

Evaluation

Chat Experience

coming soon...

Metrics

coming soon...

Citations

Cite SPIN as:

@misc{chen2024selfplayfinetuningconvertsweak,
      title={Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models}, 
      author={Zixiang Chen and Yihe Deng and Huizhuo Yuan and Kaixuan Ji and Quanquan Gu},
      year={2024},
      eprint={2401.01335},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2401.01335}, 
}

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}