metadata

license: mit
tags:
  - alignment-handbook
  - generated_from_trainer
base_model: google/gemma-2-9b-it
datasets:
  - princeton-nlp/gemma2-ultrafeedback-armorm
model-index:
  - name: princeton-nlp/gemma-2-9b-it-SimPO
    results: []

gemma-2-9b-it-SimPO Model Card

SimPO (Simple Preference Optimization) is an offline preference optimization algorithm designed to enhance the training of large language models (LLMs) with preference optimization datasets. SimPO aligns the reward function with the generation likelihood, eliminating the need for a reference model and incorporating a target reward margin to boost performance. Please refer to our preprint and github repo for more details.

Model Details

Model Description

We fine-tuned google/gemma-2-9b-it on princeton-nlp/gemma2-ultrafeedback-armorm with the SimPO objective.

Developed by: Yu Meng, Mengzhou Xia, Danqi Chen
Model type: Causal Language Model
License: gemma
Finetuned from model: google/gemma-2-9b-it

Model Sources

Repository: https://github.com/princeton-nlp/SimPO
Paper: https://arxiv.org/pdf/2405.14734

How to Get Started with the Model

import torch
from transformers import pipeline

model_id = "princeton-nlp/gemma-2-9b-it-SimPO"

generator = pipeline(
    "text-generation",
    model=model_id,
    model_kwargs={"torch_dtype": torch.bfloat16},
    device="cuda",
)
outputs = generator([{"role": "user", "content": "What's the difference between llamas and alpacas?"}],
                      do_sample=False,
                      eos_token_id=[generator.tokenizer.convert_tokens_to_ids("<end_of_turn>"), generator.tokenizer.eos_token_id],
                      max_new_tokens=200)
print(outputs[0]['generated_text'])

Training Details

Training Data

We use princeton-nlp/gemma2-ultrafeedback-armorm as the preference optimization dataset.

Training Hyperparameters

The hyperparameters used can be found in the training script.

Speeds, Sizes, Times

Fine-tuning the google/gemma-2-9b-it on princeton-nlp/gemma2-ultrafeedback-armorm takes around 100 mins to finish on 8xH100 GPUs.

Evaluation Results

models	AE2 LC	AE2 WR	AE2 Length	AH	AH Length	GSM	GSM Length	MMLU	MMLU Length
google/gemma-2-9b-it	51.1	38.1	1571	40.8	545	87.4	395	72.7	515
princeton-nlp/gemma-2-9b-it-DPO	67.8	65.4	2016	58.9	717	88.5	392	72.2	624
princeton-nlp/gemma-2-9b-it-SimPO	72.4	65.9	1833	59.1	693	88.0	341	72.2	441

Technical Specifications

Model Architecture and Objective

The model architecture is based on google/gemma-2-9b-it. We use the SimPO training objective proposed in our preprint.

Hardware

We used 8xH100 GPUs for model training.

Software

Training was done using the alignment-handbook library.

Citation

gemma model:

@article{gemma_2024,
    title={Gemma},
    url={https://www.kaggle.com/m/3301},
    DOI={10.34740/KAGGLE/M/3301},
    publisher={Kaggle},
    author={Gemma Team},
    year={2024}
}

SimPO paper:

@article{meng2024simpo,
  title={{SimPO}: Simple preference optimization with a reference-free reward},
  author={Meng, Yu and Xia, Mengzhou and Chen, Danqi},
  journal={arXiv preprint arXiv:2405.14734},
  year={2024}
}

UltraFeedback paper:

@article{cui2023ultrafeedback,
  title={{UltraFeedback}: Boosting language models with high-quality feedback},
  author={Cui, Ganqu and Yuan, Lifan and Ding, Ning and Yao, Guanming and Zhu, Wei and Ni, Yuan and Xie, Guotong and Liu, Zhiyuan and Sun, Maosong},
  journal={arXiv preprint arXiv:2310.01377},
  year={2023}
}

ArmoRM paper:

@article{wang2024interpretable,
  title={Interpretable Preferences via Multi-Objective Reward Modeling and Mixture-of-Experts},
  author={Wang, Haoxiang and Xiong, Wei and Xie, Tengyang and Zhao, Han and Zhang, Tong},
  journal={arXiv preprint arXiv:2406.12845},
  year={2024}
}

Open Portuguese LLM Leaderboard Evaluation Results

Detailed results can be found here and on the 🚀 Open Portuguese LLM Leaderboard

Metric	Value
Average	73.28
ENEM Challenge (No Images)	75.09
BLUEX (No Images)	65.37
OAB Exams	54.21
Assin2 RTE	93.82
Assin2 STS	77.82
FaQuAD NLI	70.45
HateBR Binary	89.76
PT Hate Speech Binary	66.68
tweetSentBR	66.28