MMCV1. 云服务器 AotuDL1.1 选择版本软件版本来源torch2.0.0cu118pypimmcv2.2.0pypi1.2 安装 mim 工具pip install -U openmim1.3 一键安装 MMCVmim install mmcv2.2.01.4 audodl 学术加速source /etc/network_turbo1.5 验证代理输出http_proxy、https_proxy即成功。env | grep proxy2. Windows50 系显卡2.1 PyTorch 版本软件版本来源torch2.8.0cu128pypitorchaudio2.8.0cu128pypitorchvision0.23.0cu128pypi2.2 克隆 MMCVgit clone https://github.com/open-mmlab/mmcv.git -b v2.1.02.3 安装 CUDA 12.9CUDA Toolkit 12.9 Downloads | NVIDIA Developer2.4 下载 VS 桌面开发2019 版本2022 有问题。百度网盘 请输入提取码2.5 下载并编译 MMCVpip install -U openmim cd mmcv pip install --upgrade setuptools wheel pip install -r requirements/optional.txt pip install setuptools60.2.0 pip install . -v --no-build-isolation #需要c python .dev_scripts/check_installation.py3. 本地服务器3.1 安装 CUDA Toolkit 12.8wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-ubuntu2204.pin sudo mv cuda-ubuntu2204.pin /etc/apt/preferences.d/cuda-repository-pin-600 ​ wget https://developer.download.nvidia.com/compute/cuda/12.8.0/local_installers/cuda-repo-ubuntu2204-12-8-local_12.8.0-570.86.10-1_amd64.deb sudo dpkg -i cuda-repo-ubuntu2204-12-8-local_12.8.0-570.86.10-1_amd64.deb ​ sudo cp /var/cuda-repo-ubuntu2204-12-8-local/cuda-*-keyring.gpg /usr/share/keyrings/ sudo apt-get update sudo apt-get -y install cuda-toolkit-12-83.2 配置 CUDA 环境变量打开配置文件vim ~/.bashrc在最后添加export CUDA_HOME/usr/local/cuda-12.8 export PATH$CUDA_HOME/bin:$PATH export LD_LIBRARY_PATH$CUDA_HOME/lib64:$LD_LIBRARY_PATH使配置生效source ~/.bashrc3.3 配置编译器export CCgcc export CXXg3.4 编译并验证 MMCVpip install -U openmim pip install . -v --no-build-isolation python .dev_scripts/check_installation.pyMMPRE1. Windows 环境配置主要针对 50 系显卡1.1 创建 conda 环境下载 Miniconda 或者 Anaconda。conda create -n mmpre python3.101.2 下载 PyTorch下载地址https://download.pytorch.org/whl/cu128下载对应版本的torch、torchvision、torchaudio。笔者推荐以下版本如果要使用 OpenMMLab 其他库torch-2.8.0cu128-cp310-cp310-win_amd64.whl torchaudio-2.8.0cu128-cp310-cp310-win_amd64.whl torchvision-0.23.0cu128-cp310-cp310-win_amd64.whl1.3 安装 MMCV注意要使用 2.1.0 版本。git clone https://github.com/open-mmlab/mmcv.git -b v2.1.01.4 安装 CUDA 12.8版本差不多都能用选择自己的版本。https://developer.nvidia.com/cuda-12-8-0-download-archive?target_osWindowstarget_archx86_64target_version11target_typeexe_local1.5 下载 VS 桌面开发百度网盘 请输入提取码下载使用 C 的桌面开发。注意使用setuptools60.2.0高版本不行。1.6 编译 MMCVpip install -U openmim ​ cd mmcv # 切换到刚才git clone的mmcv目录 ​ pip install --upgrade setuptools wheel # 这里会版本报错 等会解决 pip install -r requirements/optional.txt pip install setuptools60.2.0 # 解决报错 ​ pip install . -v --no-build-isolation #需要c刚才下载的桌面开发关闭 “构建隔离环境”编译时警告可以不予理会只要能编译完成 ​ python .dev_scripts/check_installation.py # 验证结果验证输出(mmpre_test) PS D:\resource\code\python\ai\mmcv python .dev_scripts/check_installation.py Start checking the installation of mmcv ... CPU ops were compiled successfully. CUDA ops were compiled successfully. mmcv has been installed successfully.1.7 安装 MMPreTraingit clone https://github.com/open-mmlab/mmpretrain.git cd mmpretrain pip install -U openmim mim install -e .运行测试python demo/image_demo.py demo/demo.JPEG configs/resnet/resnet18_8xb32_in1k.py --device cpu验证输出(mmpre_test) PS D:\resource\code\python\ai\mmpretrain python demo/image_demo.py demo/demo.JPEG configs/resnet/resnet18_8xb32_in1k.py --device cpu Inference ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% 0:00:00 { pred_label: 621, pred_score: 0.0011044980492442846, pred_class: lawn mower, mower }2. AotuDL 快速配置 MMCV 和 MMPreTrain服务器选择torch 2.0.02.1 安装 mim 工具pip install -U openmim2.2 一键安装 MMCVmim install mmcv2.2.02.3 audodl 学术加速source /etc/network_turbo2.4 安装 MMPreTraingit clone https://github.com/open-mmlab/mmpretrain.git cd mmpretrain pip install -U openmim mim install -e .2.5 验证安装python demo/image_demo.py demo/demo.JPEG configs/resnet/resnet18_8xb32_in1k.py --device cpuMMSegmentation1. 服务器1.1 克隆项目git clone https://github.com/open-mmlab/mmsegmentation.git cd mmsegmentation pip install -v -e .1.2 下载测试配置和模型mim download mmsegmentation \ --config pspnet_r50-d8_4xb2-40k_cityscapes-512x1024 \ --dest .1.3 验证推理python demo/image_demo.py \ demo/demo.png \ configs/pspnet/pspnet_r50-d8_4xb2-40k_cityscapes-512x1024.py \ pspnet_r50-d8_512x1024_40k_cityscapes_20200605_003338-2966598c.pth \ --device cuda:0 \ --out-file result.jpg1.4 常见报错报错一ModuleNotFoundError: No module named ftfy解决方法pip install ftfy报错二ModuleNotFoundError: No module named regex解决方法pip install regex -y报错三KeyError: XXDataset is not in the mmseg::dataset registry Please check解决方法python setup.py install pip install -v -e .或者python setup.py develop1.5 环境版本软件版本安装类型mmcv2.2.0pypimmengine0.10.7pypimmsegmentation1.2.2develop1.6 训练命令python train.py \ --model wtconvnext_tiny \ --drop-path 0.1 \ --data-dir /root/data/imagenet100 \ --epochs 300 \ --warmup-epochs 20 \ --batch-size 64 \ --grad-accum-steps 64 \ --sched-on-updates \ --lr 4e-3 \ --weight-decay 5e-2 \ --opt adamw \ --layer-decay 1.0 \ --aa rand-m9-mstd0.5-inc1 \ --reprob 0.25 \ --mixup 0.8 \ --cutmix 1.0 \ --model-ema \ --model-ema-decay 0.9999 \ --output checkpoints/wtconvnext_tiny_300/2. Windows2.1 安装 MMSegmentationmim install mmsegmentation项目地址GitHub - open-mmlab/mmsegmentation: OpenMMLab Semantic Segmentation Toolbox and Benchmark. · GitHub2.2 开发模式安装pip install -v -e .2.3 添加模块后重新安装python setup.py installMMDetectionWindows1. 安装项目pip install -U openmim mim install mmdet ​ git clone https://github.com/open-mmlab/mmdetection.git cd mmdetection ​ mim download mmdet \ --config rtmdet_tiny_8xb32-300e_coco \ --dest . ​ py setup.py install2. 修改 checkpoint.py从以下目录找到checkpoint.pyC:\Users\你的用户名\miniconda3\envs\hw5\lib\site-packages\mmengine\runner\修改第 347 行为checkpoint torch.load( filename, map_locationmap_location, weights_onlyFalse )3. 测试python demo/image_demo.py \ demo/demo.jpg \ rtmdet_tiny_8xb32-300e_coco.py \ --weights rtmdet_tiny_8xb32-300e_coco_20220902_112414-78e30dcc.pth \ --device cuda:04. 参考链接https://blog.csdn.net/jaydeng666/article/details/156494793?ops_request_miscelastic_search_miscrequest_idb2701dbe51cd880ce37344cf29abc393biz_id0utm_mediumdistribute.pc_search_result.none-task-blog-2allElasticSearch~search_v2-8-156494793-null-null.142v102pc_search_result_base9utm_termmmdet3d%2050%E7%B3%BB%E6%98%BE%E5%8D%A1%20windowsspm1018.2226.3001.4187MMDetection3D1. 安装依赖1.1 安装 MMDetectionpip install mmdet3.3.01.2 安装 MMSegmentationpip install mmsegmentation1.2.21.3 安装 MMDetection3Dgit clone https://github.com/open-mmlab/mmdetection3d.git cd mmdetection3d pip install -v -e . # or python setup.py develop2. 测试2.1 下载权重https://download.openmmlab.com/mmdetection3d/v1.0.0_models/votenet/votenet_16x8_sunrgbd-3d-10class/votenet_16x8_sunrgbd-3d-10class_20210820_162823-bf11f014.pth2.2 运行点云检测示例python demo/pcd_demo.py \ demo/data/sunrgbd/000017.bin \ configs/votenet/votenet_8xb16_sunrgbd-3d.py \ checkpoints/votenet_16x8_sunrgbd-3d-10class_20210820_162823-bf11f014.pth