SenseVoice-语音ASR模型权重docker镜像构建

代码示例gitcode地址

https://gitcode.com/yunting/CangjieCjoyWebSocketDemo.git

Dockerfile

FROM busybox:latest
COPY iic /models/iic

下载模型权重

# 创建python 环境
conda create -n testdemo python=3.11
# 激活环境 testdemo
conda activate testdemo
# 下载项目
git clone https://gitee.com/howardji/SenseVoice
# 进入项目目录
cd SenseVoice
# 安装依赖
pip install -r requirements.txt
#启动
#GPU
#export SENSEVOICE_DEVICE=cuda:0
#CPU
export SENSEVOICE_DEVICE=cpu
uvicorn api:app  --host 0.0.0.0 --port 50000

#模型权重目录
/Users/xxx/.cache/modelscope/hub/models/iic

按图示存放

docker构建

构建基础镜像

docker build -t sensevoice-model:1.0.0 .

查看构建的镜像

docker images | grep sensevoice-model

进入容器验证环境

docker run -it sensevoice-model:1.0.0  /bin/bash

上传镜像

• 注册账号

# 登录 测试账号 ting567 登录了，这步可省略
docker login -u ting567
# 打标签
docker tag  sensevoice-model:1.0.0 ting567/sensevoice-model:1.0.0
# 推送
docker push ting567/sensevoice-model:1.0.0