深度学习pytorch模型docker打包运行

记录工信部医疗影像评测相关代码。
公共仓库镜像拉取
docker pull ai-hub.3incloud.com/library/pytorch/pytorch:1.2-cuda10.0-cudnn7-runtime
打包
docker run -it (image ID) /bin/bash
pip install torch1.7.0+cu101 torchvision0.8.1+cu101 torchaudio==0.7.0 -f https://download.pytorch.org/whl/torch_stable.html
pip install albumentations opencv-contrib-python segmentation-models-pytorch
拷贝
docker cp (pwd) (container ID):/workspace/
bash文件
docker commit (container ID) ai-hub.3incloud.com/comp_18/eye
docker tag (image:1.0) ai-hub.3incloud.com/username/official
上传私有镜像仓库
docker login ai-hub.3incloud.com
docker push ai-hub.3incloud.com/comp_18/eye
沙箱
docker pull docker pull ai-hub.3incloud.com/username/classify:latest
docker run --gpus all -it -v /data/data1/眼表疾病数据:/workspace/data_eye ai-hub.3incloud.com/comp_18/eye /bin/bash
sudo docker cp (container ID):/workspace/pipeline/result.csv /data/result.csv
sftp上传
sftp comp_18@ai-download.3incloud.com
cd sandbox_result/
put ./result.csv ./

删除镜像
docker rmi (image ID)/$(dockerimages-q)-f删除容器dockerrm(containerID)/$(docker ps -aq)
停止/开始/重启容器
docker stop/start/restart (container ID)
不关闭退出容器
Ctrl+P+Q
进入运行中的容器
docker attach

docker19支持运行GPU环境
安装Nvidia-container-runtime

curl -s -L https://nvidia.github.io/nvidia-container-runtime/gpgkey | \
  sudo apt-key add -
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-container-runtime/$distribution/nvidia-container-runtime.list | \
  sudo tee /etc/apt/sources.list.d/nvidia-container-runtime.list
sudo apt-get update
sudo apt nvidia-container-runtime

运行命令

docker run --gpus all -it -v /data/data1/眼表疾病数据:/workspace/data_eye ai-hub.3incloud.com/comp_18/eye /bin/bash

python classify_dataset_new.py --data_dir ***** --save_csv “result.csv” --device “gpu”
Nvidia-container-runtime