133 lines
4.2 KiB
Markdown
133 lines
4.2 KiB
Markdown
<p align="center">
|
||
<img src="https://github.com/josStorer/RWKV-Runner/assets/13366013/d24834b0-265d-45f5-93c0-fac1e19562af">
|
||
</p>
|
||
|
||
<h1 align="center">RWKV Runner</h1>
|
||
|
||
<div align="center">
|
||
|
||
本项目旨在消除大语言模型的使用门槛，全自动为你处理一切，你只需要一个仅仅几MB的可执行程序。此外本项目提供了与OpenAI
|
||
API兼容的接口，这意味着一切ChatGPT客户端都是RWKV客户端。
|
||
|
||
[![license][license-image]][license-url]
|
||
[![release][release-image]][release-url]
|
||
|
||
[English](README.md) | 简体中文
|
||
|
||
### 安装
|
||
|
||
[![Windows][Windows-image]][Windows-url]
|
||
[![MacOS][MacOS-image]][MacOS-url]
|
||
[![Linux][Linux-image]][Linux-url]
|
||
|
||
[视频演示](https://www.bilibili.com/video/BV1hM4y1v76R) | [疑难解答](https://www.bilibili.com/read/cv23921171) | [预览](#Preview) | [下载][download-url] | [懒人包](https://pan.baidu.com/s/1wchIUHgne3gncIiLIeKBEQ?pwd=1111)
|
||
|
||
[license-image]: http://img.shields.io/badge/license-MIT-blue.svg
|
||
|
||
[license-url]: https://github.com/josStorer/RWKV-Runner/blob/master/LICENSE
|
||
|
||
[release-image]: https://img.shields.io/github/release/josStorer/RWKV-Runner.svg
|
||
|
||
[release-url]: https://github.com/josStorer/RWKV-Runner/releases/latest
|
||
|
||
[download-url]: https://github.com/josStorer/RWKV-Runner/releases
|
||
|
||
[Windows-image]: https://img.shields.io/badge/-Windows-blue?logo=windows
|
||
|
||
[Windows-url]: https://github.com/josStorer/RWKV-Runner/blob/master/build/windows/Readme_Install.txt
|
||
|
||
[MacOS-image]: https://img.shields.io/badge/-MacOS-black?logo=apple
|
||
|
||
[MacOS-url]: https://github.com/josStorer/RWKV-Runner/blob/master/build/darwin/Readme_Install.txt
|
||
|
||
[Linux-image]: https://img.shields.io/badge/-Linux-black?logo=linux
|
||
|
||
[Linux-url]: https://github.com/josStorer/RWKV-Runner/blob/master/build/linux/Readme_Install.txt
|
||
|
||
</div>
|
||
|
||
#### 注意 目前RWKV中文模型质量一般，推荐使用英文模型体验实际RWKV能力
|
||
|
||
#### 对于Mac和Linux用户，请手动安装 Python3.10 (通常最新的系统已经内置了)
|
||
|
||
#### 预设配置已经开启自定义CUDA算子加速，速度更快，且显存消耗更少。如果你遇到可能的兼容性问题，前往配置页面，关闭`使用自定义CUDA算子加速`
|
||
|
||
#### 对于不同的任务，调整API参数会获得更好的效果，例如对于翻译任务，你可以尝试设置Temperature为1，Top_P为0.3
|
||
|
||
## 功能
|
||
|
||
- RWKV模型管理，一键启动
|
||
- 与OpenAI API完全兼容，一切ChatGPT客户端，都是RWKV客户端。启动模型后，打开 http://127.0.0.1:8000/docs 查看详细内容
|
||
- 全自动依赖安装，你只需要一个轻巧的可执行程序
|
||
- 预设了2G至32G显存的配置，几乎在各种电脑上工作良好
|
||
- 自带用户友好的聊天和补全交互页面
|
||
- 易于理解和操作的参数配置
|
||
- 内置模型转换工具
|
||
- 内置下载管理和远程模型检视
|
||
- 多语言本地化
|
||
- 主题切换
|
||
- 自动更新
|
||
|
||
## API并发压力测试
|
||
|
||
```bash
|
||
ab -p body.json -T application/json -c 20 -n 100 -l http://127.0.0.1:8000/chat/completions
|
||
```
|
||
|
||
body.json:
|
||
|
||
```json
|
||
{
|
||
"messages": [
|
||
{
|
||
"role": "user",
|
||
"content": "Hello"
|
||
}
|
||
]
|
||
}
|
||
```
|
||
|
||
## Todo
|
||
|
||
- [ ] 模型训练功能
|
||
- [x] CUDA算子int8提速
|
||
- [x] macOS支持
|
||
- [x] linux支持
|
||
- [ ] 本地状态缓存数据库
|
||
|
||
## 相关仓库:
|
||
|
||
- RWKV-4-Raven: https://huggingface.co/BlinkDL/rwkv-4-raven/tree/main
|
||
- ChatRWKV: https://github.com/BlinkDL/ChatRWKV
|
||
- RWKV-LM: https://github.com/BlinkDL/RWKV-LM
|
||
|
||
## Preview
|
||
|
||
### 主页
|
||
|
||
|
||
|
||
### 聊天
|
||
|
||
|
||
|
||
### 补全
|
||
|
||
|
||
|
||
### 配置
|
||
|
||
|
||
|
||
### 模型管理
|
||
|
||
|
||
|
||
### 下载管理
|
||
|
||
|
||
|
||
### 设置
|
||
|
||
|