腾讯云为GPU云服务器GC3vq机型提供特定驱动安装脚本,同时提供CUDA、cuDNN和相关的AI框架自动安装脚本,在活动页购买的机器,腾讯云提供以下三种脚本部署,您可以根据需要选择:
机型 | 操作系统 | 软件环境 | 执行命令 |
---|---|---|---|
GC3vq | Ubuntu 18.04/20.04 CentOS 7.6/7.8/7.9 | 【驱动安装】vqGPU-DRIVER1.0.0_cuda11.4.3:CUDA11.4.3 cuDNN8.2.4 vqGPU驱动 | wget https://gpu-related-scripts-1251783334.cos.ap-guangzhou.myzijiebao.com/gpu-auto-install/gpu_auto_install.sh && wget https://gpu-related-scripts-1251783334.cos.ap-guangzhou.myzijiebao.com/gpu-auto-install/vqGPU-DRIVER1.0.0_cuda11.4.3.txt && sudo bash ./gpu_auto_install.sh install --config_file=./vqGPU-DRIVER1.0.0_cuda11.4.3.txt && source /etc/*bashrc && source ${HOME}/.bashrc |
【AI环境】tf2.8.0_vqGPU-DRIVER1.0.0_cuda11.4.3:TensorFlow 2.8.0 Miniconda OpenCV 4 Python 3.9.12 CUDA11.4.2 cuDNN8.2.4 vqGPU驱动 | wget https://gpu-related-scripts-1251783334.cos.ap-guangzhou.myzijiebao.com/gpu-auto-install/gpu_auto_install.sh && wget https://gpu-related-scripts-1251783334.cos.ap-guangzhou.myzijiebao.com/gpu-auto-install/tf2.8.0_vqGPU-DRIVER1.0.0_cuda11.4.3.txt && sudo bash ./gpu_auto_install.sh install --config_file=./tf2.8.0_vqGPU-DRIVER1.0.0_cuda11.4.3.txt && source /etc/*bashrc && source ${HOME}/.bashrc | ||
【AI环境】pt1.9.1_vqGPU-DRIVER1.0.0_cuda11.4.3:Pytorch 1.9.1 torchvision0.10.0 Miniconda OpenCV 4 Python 3.8 CUDA11.4.2 cuDNN8.2.4 vqGPU驱动 | wget https://gpu-related-scripts-1251783334.cos.ap-guangzhou.myzijiebao.com/gpu-auto-install/gpu_auto_install.sh && wget https://gpu-related-scripts-1251783334.cos.ap-guangzhou.myzijiebao.com/gpu-auto-install/pt1.9.1_vqGPU-DRIVER1.0.0_cuda11.4.3.txt && sudo bash ./gpu_auto_install.sh install --config_file=./pt1.9.1_vqGPU-DRIVER1.0.0_cuda11.4.3.txt && source /etc/*bashrc && source ${HOME}/.bashrc |
使用流程
1、进入云服务器控制台,进入对应机器主页。
2、设置安全组策略,放通8888端口
(1)点击“安全组”-“规则预览”-“编辑规则”
(2)需要在“入站规则”和“出站规则”都添加8888端口放通策略
(3)分别在“入站规则”和“出站规则”栏下点击“添加规则”,按照下图进行配置
3、登陆机器,使用脚本部署Miniconda AI环境
如果您需要 TensorFlow 2.8.0,登录子机执行下列命令:
代码语言:javascript复制wget https://gpu-related-scripts-1251783334.cos.ap-guangzhou.myzijiebao.com/gpu-auto-install/gpu_auto_install.sh && wget https://gpu-related-scripts-1251783334.cos.ap-guangzhou.myzijiebao.com/gpu-auto-install/tf2.8.0_vqGPU-DRIVER1.0.0_cuda11.4.3.txt && sudo bash ./gpu_auto_install.sh install --config_file=./tf2.8.0_vqGPU-DRIVER1.0.0_cuda11.4.3.txt && (source /etc/bash.bashrc || source ${HOME}/.bashrc)
如果您需要 Pytorch 1.9.1 ,登录子机执行下列命令:
代码语言:javascript复制wget https://gpu-related-scripts-1251783334.cos.ap-guangzhou.myzijiebao.com/gpu-auto-install/gpu_auto_install.sh && wget https://gpu-related-scripts-1251783334.cos.ap-guangzhou.myzijiebao.com/gpu-auto-install/pt1.9.1_vqGPU-DRIVER1.0.0_cuda11.4.3.txt && sudo bash ./gpu_auto_install.sh install --config_file=./pt1.9.1_vqGPU-DRIVER1.0.0_cuda11.4.3.txt && (source /etc/bash.bashrc || source ${HOME}/.bashrc)
安装完成后可以看到JupyterNotebook访问链接,复制链接即可访问:
【说明】 *如果重启,token会更新,可使用jupyter notebook list查看token。 *如果希望使用密码登录jupyter notebook,可按如下步骤执行: (1)执行sudo jupyter notebook password更改密码; (2)执行sudo systemctl restart jupyter.service 重启jupyter notebook服务。