---
base_model:
- hfl/llama-3-chinese-8b-instruct-v3
- UnicomLLM/Unichat-llama3-Chinese-8B
- shenzhi-wang/Llama3-8B-Chinese-Chat
- yaojialzc/Gigi-Llama3-8B-Chinese-zh
- Rookie/Llama-3-8B-Instruct-Chinese
- FlagAlpha/Llama3-Chinese-8B-Instruct
library_name: transformers
tags:
- mergekit
- merge
license: llama3
language:
- en
- zh
---
# Llama3-zhcn

<details>
  <summary>English</summary>

  # A Merged Llama 3 Model for Enhanced Chinese Understanding
This model is a merge of several pre-trained Llama 3 8B language models specifically focused on Simplified Chinese (China), created using [mergekit](https://github.com/cg123/mergekit).

## Purpose

Llama3-zhcn aims to deliver a Llama 3 model with deep Chinese cultural, historical, and linguistic comprehension. This model serves dual purposes: handling diverse everyday tasks and providing a solid foundation for additional merging and fine-tuning. As Chinese model development trends away from the Llama series, this merge strives to maintain and improve Llama 3's Chinese language capabilities

## Limitations

*   **No Vision Capabilities:** This model is based on Llama 3 and does not include vision capabilities found in some later models like 3.1 and 3.2.
*   **Historical Accuracy:** While the model possesses a good understanding of Chinese history, fact-checking is still recommended to ensure accuracy.
*   **Translation and Revision:** The model is capable of performing translations and revisions in English and Chinese. However, optimal results may require prompt engineering.

## Merge Details

### Merge Method

This model was created using the [Linear](https://arxiv.org/abs/2203.05482) merge method, utilizing the Meta-Llama-3-8B-Instruct tokenizer.

### Models Merged

The following models were included in the merge:

*   [hfl/llama-3-chinese-8b-instruct-v3](https://huggingface.co/hfl/llama-3-chinese-8b-instruct-v3)
*   [UnicomLLM/Unichat-llama3-Chinese-8B](https://huggingface.co/UnicomLLM/Unichat-llama3-Chinese-8B)
*   [shenzhi-wang/Llama3-8B-Chinese-Chat](https://huggingface.co/shenzhi-wang/Llama3-8B-Chinese-Chat)
*   [yaojialzc/Gigi-Llama3-8B-Chinese-zh](https://huggingface.co/yaojialzc/Gigi-Llama3-8B-Chinese-zh)
*   [Rookie/Llama-3-8B-Instruct-Chinese](https://huggingface.co/Rookie/Llama-3-8B-Instruct-Chinese)
*   [FlagAlpha/Llama3-Chinese-8B-Instruct](https://huggingface.co/FlagAlpha/Llama3-Chinese-8B-Instruct)
</details>

<details>
  <summary>Chinese</summary>
  
  # 增强中文理解的合并Llama 3模型

该模型是针对简体中文（中国）的多个预训练了8B语言模式进行融合，使用[mergekit](https://github.com/cg123/mergekit)创建。

## 目标：

Llama3-zhcn旨在提供具有深入的中华文化、历史和语法理解能力。该模型有两个目的：处理各种日常任务，并为进一步组合或微调奠定坚实基础。在中国语言模式发展趋势从Llama系列转向时，这个融合尝试维护并改进Llama 3的中文语言功能。

## 限制：

*   **无视觉能力：**该模型基于Llama 3，不包括一些后续版本如3.1和3.2中具有的一些图像处理能力。
*   **历史准确性：**虽然这个模型对中国有良好的理解，但仍然建议进行事实核查以保证精度。
*   **翻译与修订：**该模型可以在英语和中文之间执行翻译和修改。然而，最佳结果可能需要引导工程。

## 合并细节：

### 组合方法

使用线性（Linear）组合法创建此模型，并利用Meta-Llama-3-8B-Instruct分词器进行处理。

### 被融入的模型

以下是被包含在内的模型：
*   [hfl/llama-3-chinese-8b-instruct-v3](https://huggingface.co/hfl/llama-3-chinese-8b-instruct-v3)
*   [UnicomLLM/Unichat-llama3-Chinese-8B](https://huggingface.co/UnicomLLM/Unichat-llama3-Chinese-8B)
*   [shenzhi-wang/Llama3-8B-Chinese-Chat](https://huggingface.co/shenzhi-wang/Llama3-8B-Chinese-Chat)
*   [yaojialzc/Gigi-Llama3-8B-Chinese-zh](https://huggingface.co/yaojialzc/Gigi-Llama3-8B-Chinese-zh)
*   [Rookie/Llama-3-8B-Instruct-Chinese](https://huggingface.co/Rookie/Llama-3-8B-Instruct-Chinese)
*   [FlagAlpha/Llama3-Chinese-8B-Instruct](https://huggingface.co/FlagAlpha/Llama3-Chinese-8B-Instruct)
</details>