Quantization made by Richard Erkhov.

Mistral7B-PairRM-SPPO - GGUF

Model creator: https://huggingface.co/UCLA-AGI/
Original model: https://huggingface.co/UCLA-AGI/Mistral7B-PairRM-SPPO/

Name	Quant method	Size
Mistral7B-PairRM-SPPO.Q2_K.gguf	Q2_K	2.53GB
Mistral7B-PairRM-SPPO.Q3_K_S.gguf	Q3_K_S	2.95GB
Mistral7B-PairRM-SPPO.Q3_K.gguf	Q3_K	3.28GB
Mistral7B-PairRM-SPPO.Q3_K_M.gguf	Q3_K_M	3.28GB
Mistral7B-PairRM-SPPO.Q3_K_L.gguf	Q3_K_L	3.56GB
Mistral7B-PairRM-SPPO.IQ4_XS.gguf	IQ4_XS	3.67GB
Mistral7B-PairRM-SPPO.Q4_0.gguf	Q4_0	3.83GB
Mistral7B-PairRM-SPPO.IQ4_NL.gguf	IQ4_NL	3.87GB
Mistral7B-PairRM-SPPO.Q4_K_S.gguf	Q4_K_S	3.86GB
Mistral7B-PairRM-SPPO.Q4_K.gguf	Q4_K	4.07GB
Mistral7B-PairRM-SPPO.Q4_K_M.gguf	Q4_K_M	4.07GB
Mistral7B-PairRM-SPPO.Q4_1.gguf	Q4_1	4.24GB
Mistral7B-PairRM-SPPO.Q5_0.gguf	Q5_0	4.65GB
Mistral7B-PairRM-SPPO.Q5_K_S.gguf	Q5_K_S	4.65GB
Mistral7B-PairRM-SPPO.Q5_K.gguf	Q5_K	4.78GB
Mistral7B-PairRM-SPPO.Q5_K_M.gguf	Q5_K_M	4.78GB
Mistral7B-PairRM-SPPO.Q5_1.gguf	Q5_1	5.07GB
Mistral7B-PairRM-SPPO.Q6_K.gguf	Q6_K	5.53GB
Mistral7B-PairRM-SPPO.Q8_0.gguf	Q8_0	7.17GB

Original model description:

license: apache-2.0 datasets: - openbmb/UltraFeedback language: - en pipeline_tag: text-generation

Self-Play Preference Optimization for Language Model Alignment (https://arxiv.org/abs/2405.00675)

Mistral7B-PairRM-SPPO

This model was developed using Self-Play Preference Optimization, based on the mistralai/Mistral-7B-Instruct-v0.2 architecture as starting point. We utilized the prompt sets from the openbmb/UltraFeedback dataset, splited to 3 parts for 3 iterations by snorkelai/Snorkel-Mistral-PairRM-DPO-Dataset. All responses used are synthetic.

While K = 5, this model uses three samples to estimate the soft probabilities P(y_w > y_l) and P(y_l > y_w). These samples include the winner, the loser, and another random sample. This approach has shown to deliver better performance on AlpacaEval 2.0 compared to the results reported in our paper.

❗Please refer to the original checkpoint at UCLA-AGI/Mistral7B-PairRM-SPPO-Iter3 as reported in our paper. We anticipate that the version in the paper demonstrates a more consistent performance improvement across all evaluation tasks.

Links to Other Models

Model Description

Model type: A 7B parameter GPT-like model fine-tuned on synthetic datasets.
Language(s) (NLP): Primarily English
License: Apache-2.0
Finetuned from model: mistralai/Mistral-7B-Instruct-v0.2

AlpacaEval Leaderboard Evaluation Results

Model	LC. Win Rate	Win Rate	Avg. Length
Mistral7B-PairRM-SPPO	30.46	32.14	2114

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-07
eta: 1000
per_device_train_batch_size: 8
gradient_accumulation_steps: 1
seed: 42
distributed_type: deepspeed_zero3
num_devices: 8
optimizer: RMSProp
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_train_epochs: 18.0 (stop at epoch=1.0)

Citation

@misc{wu2024self,
      title={Self-Play Preference Optimization for Language Model Alignment}, 
      author={Wu, Yue and Sun, Zhiqing and Yuan, Huizhuo and Ji, Kaixuan and Yang, Yiming and Gu, Quanquan},
      year={2024},
      eprint={2405.00675},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}