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="qgallouedec/xpo-qwen2", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=500)[0]
print(output["generated_text"][1]["content"])

Training procedure

Visualize in Weights & Biases

This model was trained with XPO, a method introduced in Exploratory Preference Optimization: Harnessing Implicit Q*-Approximation for Sample-Efficient RLHF.

Framework versions

TRL: 0.12.0.dev0
Transformers: 4.45.0.dev0
Pytorch: 2.4.1
Datasets: 3.0.0
Tokenizers: 0.19.1

Citations

Cite XPO as:

bibtex

@article{jung2024binary,
    title        = {{Binary Classifier Optimization for Large Language Model Alignment}},
    author       = {Seungjae Jung and Gunsoo Han and Daniel Wontae Nam and Kyoung{-}Woon On},
    year         = 2024,
    eprint       = {arXiv:2404.04656}
}

Cite TRL as:

bibtex

@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}}
}