从零开始在Ubuntu上利用Docker部署FoundationPose项目
系统环境CPUIntel® Core™ i7-14650HX显卡NVIDIA Geforce RTX 4060 Laptop Driver535.288.01CUDA11.8操作系统Ubuntu 22.04.5 LTSFoundationPose项目地址https://github.com/NVlabs/FoundationPose0.准备工作确保系统满足基础要求确保Docker以及NVIDIA Container Toolkit正确安装并正常运行并保证Docker可以正确拉取镜像1.拉取镜像克隆项目到本地git clone https://github.com/NVlabs/FoundationPose.git定位到项目文件夹拉取镜像cd FoundationPose/docker请注意确认电脑的显卡是否为较新架构NVIDIA RTX 40系列及以后如果是运行sudo docker pull shingarey/foundationpose_custom_cuda121:latest sudo docker tag shingarey/foundationpose_custom_cuda121:latest foundationpose:latest如果是比较老的显卡则可以运行sudo docker pull wenbowen123/foundationpose sudo docker tag wenbowen123/foundationpose foundationpose运行容器bash run_container.sh2.安装依赖bash build_all.sh注意如果安装时报错如#error C17 or later compatible compiler is required to use ATen.需要将bundlesdf/mycuda/setup.py中的两处c14改为c17然后重新运行bash build_all.sh命令安装3.运行demo权重文件(weights)https://drive.google.com/drive/folders/1DFezOAD0oD1BblsXVxqDsl8fj0qzB82idemo数据(demo_data)https://drive.google.com/drive/folders/1pRyFmxYXmAnpku7nGRioZaKrVJtIsroP里面的文件全部下载解压权重文件放到weights/里数据放到demo_data里然后直接运行demopython run_demo.py启动项目再次启动项目时运行sudo docker start foundationpose docker exec -it foundationpose bash即可
本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如若转载,请注明出处:http://www.coloradmin.cn/o/2573304.html
如若内容造成侵权/违法违规/事实不符,请联系多彩编程网进行投诉反馈,一经查实,立即删除!