mannix / phi3-mini-cpo-simpo

Phi-3-mini-4K-instruct with CPO-SimPO

25次提取更新于6周前

6周前更新

6周前

5431e2d157ae · 2.0GB

说明文档

Phi-3-mini-4K-instruct with CPO-SimPO

使用 i-matrix calibration_datav3.txt 进行的量化
转换为 fp32 的 Saftensors

本仓库包含经过 CPO-SimPO 技术增强的 Phi-3-mini-4K-instruct 模型。CPO-SimPO 结合了对比偏好优化 (CPO) 和简单偏好优化 (SimPO)。

简介

Phi-3-mini-4K-instruct 是一种针对基于指令的任务进行优化的模型。这种方法在关键基准测试中展现了显著的改进，推动了 AI 偏好学习的发展边界。

什么是 CPO-SimPO？

CPO-SimPO 是一种新技术，结合了 CPO 和 SimPO 的元素。

对比偏好优化（CPO）: 添加了一种行为克隆正则化器，以确保模型保持接近偏好数据分布。
简单偏好优化（SimPO）: 包括长度归一化和目标奖励边际，以防止生成长但质量低下的序列。

GitHub

CPO-SIMPO

模型性能

即将推出！

关键改进

增强模型性能: 显著的分数提升，特别是GSM8K（上升8.49分！）和TruthfulQA（上升2.07分）。
质量控制: 通过长度归一化和奖励边际，提高高质量序列的生成。
平衡优化: BC正则化器有助于在保持学习偏好完整性的同时，不偏离偏好数据分布。

# Phi-3-mini-4K-instruct with CPO-SimPO

- Quantizations with i-matrix `calibration_datav3.txt`
- Saftensors converted to fp32

This repository contains the Phi-3-mini-4K-instruct model enhanced with the CPO-SimPO technique. CPO-SimPO combines Contrastive Preference Optimization (CPO) and Simple Preference Optimization (SimPO).

## Introduction

Phi-3-mini-4K-instruct is a model optimized for instruction-based tasks. This approach has demonstrated notable improvements in key benchmarks, pushing the boundaries of AI preference learning.

### What is CPO-SimPO?

CPO-SimPO is a novel technique, which combines elements from CPO and SimPO:

- **Contrastive Preference Optimization (CPO):** Adds a behavior cloning regularizer to ensure the model remains close to the preferred data distribution.
- **Simple Preference Optimization (SimPO):** Incorporates length normalization and target reward margins to prevent the generation of long but low-quality sequences.

### Github

**[CPO-SIMPO](https://github.com/fe1ixxu/CPO_SIMPO)**

## Model Performance

COMING SOON!

### Key Improvements:
- **Enhanced Model Performance:** Significant score improvements, particularly in GSM8K (up by 8.49 points!) and TruthfulQA (up by 2.07 points).
- **Quality Control:** Improved generation of high-quality sequences through length normalization and reward margins.
- **Balanced Optimization:** The BC regularizer helps maintain the integrity of learned preferences without deviating from the preferred data distribution.

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)