YOLOv5 报错：“NotImplementedError: Could not run ‘torchvision::nms‘ with arguments from the ‘CUDA‘ back
免费下载资源
是torchvisiao-cuda没装好，重装下
直接：
pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio===0.9.0 -f https://download.pytorch.org/whl/torch_stable.html -i https://pypi.douban.com/simple
文章目录
20230816
YOLOv5中的"torchvision::nms"错误问题解决方法
本文主要探讨YOLOv5在使用过程中可能遇到的一个常见错误：“NotImplementedError: Could not run ‘torchvision::nms‘ with arguments from the ‘CUDA‘ backend”。文章将从问题的产生原因、解决策略以及未来可能的优化方向进行深度剖析。
一、问题的产生原因
1.1 错误概述
“torchvision::nms”是PyTorch库中的一个函数，用于实现非极大值抑制（Non-Maximum Suppression，NMS）。NMS是物体检测算法中的一个重要步骤，它能够帮助我们去除冗余的检测框，保留最佳的预测结果。
boxes = torch.rand(5, 4) * 100
scores = torch.rand(5)
keep = torchvision.ops.nms(boxes, scores, iou_threshold)
上述错误表明，'torchvision::nms’无法在’cuda’后端运行。通常，这种问题出现在以下两种情况：
1.1.1 PyTorch与Torchvision版本不匹配
如果正在使用的PyTorch版本和Torchvision版本之间存在不兼容性，就可能会遇到这个问题。这是因为Torchvision库的某些功能依赖于特定版本的PyTorch，当两者版本不匹配时，就可能导致问题。
1.1.2 CUDA版本问题
另一种可能的原因是CUDA版本问题。具体来说，如果安装的CUDA版本与PyTorch或Torchvision所需的版本不符，也可能会出现此问题。
二、解决策略
2.1 确保PyTorch和Torchvision版本匹配
首先，应确保PyTorch和Torchvision的版本完全匹配。可以通过下面的代码查看当前的版本信息：
import torch
import torchvision
print(torch.__version__)
print(torchvision.__version__)
如果发现版本不匹配，可以通过卸载并重新安装对应版本的库来解决：
pip uninstall torch torchvision
pip install torch==1.x.x torchvision==0.y.y
其中，“1.x.x”和“0.y.y”需要替换为实际需要的版本号。
2.2 确保CUDA版本正确
其次，需要确认CUDA的版本是否与PyTorch和Torchvision兼容。可以使用以下命令查看CUDA版本：
nvcc --version
如果CUDA版本不符合要求，需要下载并安装正确版本的CUDA。
三、优化方向
尽管上述解决策略可以有效地解决问题，但还有进一步的优化空间。例如，可以考虑实现一个检查脚本，在每次运行YOLOv5之前自动检查PyTorch、Torchvision和CUDA的版本，并在版本不匹配时给出警告。
参考文章：
ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ
ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ
ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ
ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ
ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ
ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ ᅟᅠ
GitHub 加速计划 / yo / yolov5
649
37
下载
旨在为数千万中国开发者提供一个无缝且高效的云端环境，以支持学习、使用和贡献开源项目。
更多推荐
10
0- 0
已为社区贡献212条内容
所有评论(0)