https://github.com/Paraworks/vits_with_chatgpt-gpt3
https://github.com/Paraworks/vits_with_chatgpt-gpt3
Last synced: 9 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/Paraworks/vits_with_chatgpt-gpt3
- Owner: Paraworks
- License: mit
- Created: 2022-12-26T04:57:31.000Z (about 3 years ago)
- Default Branch: onnx
- Last Pushed: 2023-11-02T07:12:06.000Z (about 2 years ago)
- Last Synced: 2024-11-07T06:41:07.578Z (about 1 year ago)
- Language: Python
- Size: 24.6 MB
- Stars: 389
- Watchers: 6
- Forks: 52
- Open Issues: 4
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# [无ui版及ChatGLM部署流程](https://github.com/Paraworks/vits_with_chatgpt-gpt3/tree/window)
# 部署流程
## I:启用前端应用，从[Live2DMascot](https://github.com/Arkueid/Live2DMascot)仓库下载后，修改config.json文件
```sh
"ChatAPI" :
{
"ChatSavePath" : "chat", //聊天音频和文本保存路径
"CustomChatServer" :
{
"HostPort" : "http://yourhost:8080", //服务器地址，端口默认8080
"On" : true, //开启自定义聊天接口
"ReadTimeOut" : 114, //等待响应时间(s)
"Route" : "/chat" //路径
},
```
## Window一键部署
```sh
#前置条件 已安装Anaconda
conda create -n chatbot python=3.8
conda activate chatbot
git clone https://huggingface.co/spaces/Mahiruoshi/vits-chatbot
cd vits-chatbot
pip install -r requirements.txt
python main.py
```
# 合成日语时要安装pyopenjtalk或者编译好的日语cleaner文件(效果不一定好)，所以你完全可以选择忽视该模块的安装。
在cleaner程序中，也就是text文件下的[cleaners.py](https://github.com/Paraworks/vits_with_chatgpt-gpt3/blob/onnx/text/cleaners.py),注释掉所有的japanese模块，比如说:
```sh
#第3行
from text.japanese import japanese_to_romaji_with_accent, japanese_to_ipa, japanese_to_ipa2, japanese_to_ipa3
```
在你所采用的config.json文件中，找到对应的cleaner，比如说zh_ja_mixture_cleaners,然后注释掉这一段
```sh
#第50行开始
for japanese_text in japanese_texts:
cleaned_text = japanese_to_romaji_with_accent(
japanese_text[4:-4]).replace('ts', 'ʦ').replace('u', 'ɯ').replace('...', '…')
text = text.replace(japanese_text, cleaned_text+' ', 1)
```
## II:server端启动后端api程序(Windows也可以)
如需使用pyopenjtalk，则需要先安装好cmake, 结合自己的系统搜索相关安装教程
## Linux
```sh
#Installing cmake and FFmpeg,showing FFmpeg
sudo apt update
sudo apt upgrade
sudo apt install ffmpeg
ffmpeg -version
#Creating enviornments
conda create -n chatbot python=3.8
conda init bash
bash
conda activate chatbot
git clone https://huggingface.co/Mahiruoshi/vits_with_chatbot
cd vits_with_chatbot
pip install -r requirements.txt
# 控制面板兼启动文件
python main.py
# * Running on http://127.0.0.1:8080
# * Running on http://172.16.5.4:8080
#Running on local URL: http://127.0.0.1:7860
#端口号7860是面板，8080是api
```
(较为复杂，建议参考另一个分支部署)如需使用chatglm,需提前部署好环境[在自己的环境下安装好依赖](https://github.com/THUDM/ChatGLM-6B)。建议protobuf==3.20.0，
transformers>=4.26.1，否则会有奇怪报错
## Window详细说明
## I 安装[FFmpeg](https://zhuanlan.zhihu.com/p/118362010)并且添加环境变量
## II.安装[Torch+gpu](https://blog.csdn.net/qq_44173699/article/details/126312680)(如需cpu推理则跳过)
## III.[cmake及pyopenjtalk安装(可略过)](https://www.bilibili.com/video/BV13t4y1V7DV/?spm_id_from=333.880.my_history.page.click)
## (Alternative).使用封装版的Japanese cleaner,用该text[文件夹](https://github.com/Paraworks/vits_with_chatgpt-gpt3/tree/window/text)替换原本的text文件夹,然后从该仓库的release中[下载](https://github.com/NaruseMioShirakana/JapaneseCleaner)，将cleaners压缩包解压至vits项目的路径下
## V.python创建虚拟环境后
```sh
git clone https://huggingface.co/Mahiruoshi/vits_with_chatbot
cd vits_with_chatbot
pip install -r requirements.txt
python main.py
```
# 面板说明
## 完成chatbot方式选择及vits模型加载
可供选择的方式: gpt3.5/gpt3的api，CHATGLM
方法：将路径或者api key填写进文本框中
## 测试api是否启动
# [添加vits模型](https://github.com/Paraworks/vits_with_chatgpt-gpt3/blob/onnx/checkpoints/README.md)
# [onnx导出程序，克隆该仓库后运行](https://huggingface.co/Mahiruoshi/vits_with_chatbot/blob/main/export_onnx.py)
# 由于openai将扩大对非官方api的打压力度，如以盈利为目的，建议采用官方的api key以及openai标准库