README.md

---
license: llama3
language:
- en
- zh
datasets:
- unalignment/toxic-dpo-v0.2
- shareAI/DPO-zh-en-emoji
---

# Llama3-70B-Orion-Chinese-SE

## 概述

本模型是从[failspy/Llama-3-70B-Instruct-abliterated-v3](https://huggingface.co/failspy/Llama-3-70B-Instruct-abliterated-v3)经过DPO微调而来

相较于[Orion-zhen/Llama3-70B-Orion-Chinese](https://huggingface.co/Orion-zhen/Llama3-70B-Orion-Chinese), 本模型精简了DPO数据集, 力求对原模型影响最小的情况下获得较好的中文性能和emoji体验

本模型的emoji触发概率高于[Orion-zhen/Llama3-70B-Orion-Chinese](https://huggingface.co/Orion-zhen/Llama3-70B-Orion-Chinese), 在其他方面稍逊于后者