登录账户export DOCKER_REGISTRY=registry.cn-hangzhou.aliyuncs.com/cqy_testdocker login $DOCKER_REGISTRY --username=西交陈清源 创建自己的镜像拉取基础镜像 docker pull registry.cn-shanghai.aliyuncs.com/tcc-public/python:3#基础镜像地址#https://tianchi.aliyun.com/forum/postDetail?spm=5176.12586973.0.0.176a4127dBw2Xr&postId=67720 获取镜像工具包，根据自己的需求去修改dockerfile wget http://tianchi-media.oss-cn-beijing.aliyuncs.com/docker-tool/tianchi_docker.zipunzip tianchi_docker.zip 或者进入镜像修改 docker run --gpus all -it registry.cn-shanghai.aliyuncs.com/tcc-public/pytorch:1.4-cuda10.1-py3 /bin/bash#测试时如需数据集，则可以将本地路径映射到镜像中docker run --gpus all -it -v /tmp:/tcdata registry.cn-shanghai.aliyuncs.com/tcc-public/pytorch:1.4-cuda10.1-py3 /bin/bashpip intall ***#新开一个终端#查看当前容器（运行中的镜像）docker ps #打包生成新的base镜像docker commit 容器ID registry.cn-shenzhen.aliyuncs.com/test_for_tianchi/test_for_tianchi_submit:base 最后执行docker_build.sh或者分步执行： #构建docker build -t registry.cn-shenzhen.aliyuncs.com/test_for_tianchi/test_for_tianchi_submit:1.0 .#本地测试nvidia-docker run your_image sh run.sh#上传docker push registry.cn-shenzhen.aliyuncs.com/test_for_tianchi/test_for_tianchi_submit:1.0
下载地址:
IDEA 配置Docker的过程
docker容器间跨宿主机通信-基于overlay的实现方法 |
