Edit model card

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

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}
}
Downloads last month
106,328
Safetensors
Model size
9.24B params
Tensor type
BF16
Β·
Inference API
Input a message to start chatting with princeton-nlp/gemma-2-9b-it-SimPO.

Model tree for princeton-nlp/gemma-2-9b-it-SimPO

Base model

google/gemma-2-9b
Finetuned
(69)
this model
Finetunes
8 models
Merges
28 models
Quantizations
19 models

Dataset used to train princeton-nlp/gemma-2-9b-it-SimPO

Spaces using princeton-nlp/gemma-2-9b-it-SimPO 6

Collection including princeton-nlp/gemma-2-9b-it-SimPO