Error compiling objects for extension

[wisest-cui]

Describe the bug
when i run the commend "python setup.py develop",there is a ERROR: Error compiling objects for extension
I have check the environment and the cuda is availiable. my environment is same to what requirement.txt describe.

I ask to Google and find I shoud to replace "AT_CHECK" to "TORCH_CHECK" and I find "AT_CHECK" in some path: mmdet/opt/ /src/ .cpp

Reproduction
python setup.py develop

[zhengye1995]

What is your version of PyTorch, I used torch1.1.0 in this repo.

[wisest-cui]

thank you for your apply!
我的环境是：

• OS: [e.g., Ubuntu 16.04.6]
• GCC [e.g., 5.4.0]
• PyTorch version [e.g., 1.1.0] 1.10.0
• How you installed PyTorch [e.g. pip, conda, source]
• GPU model [e.g.,4 个 3090]
• CUDA and CUDNN version：10.0.

并且我在https://download.pytorch.org/whl/torch/ 下载pytorch轮子的地方没有找到torch1.1.0 对应有cuda的版本。不知作者您的pytorch=1.1.0 torchvision=0.3.0 cudatoolkit=10.0是如何匹配的 (按照我的理解cuda10.0的版本和pytorch1.1.0不匹配,所以我就以为这里pytorch1.1.0是您写错了 所以就下载了 1.10.0)

[zhengye1995]

I install torch through: conda install pytorch==1.1.0 torchvision==0.3.0 cudatoolkit=10.0 -c pytorch

By the way, cuda10 may not be compatible with 3090.

[wisest-cui]

我执行 nvidia-sim 命令查看显卡最高支持到11.2，应该是向下兼容的吧（自带翻译）
然后我有另外一个问题 我的cuda是通过cuda官网 下载runfile 到 usr/local/cuda 但是只通过 conda install cudatoolkit=10.0的话 通过nvcc -V是无法显示cuda下载成功的。我的理解是 conda install 主要包含应用程序在使用CUDA相关的功能时所依赖的动态链接库。conda安装只会安装一些计算库，但不会安装编译工具。
所以步骤是不是应该是先从cuda官网下载cuda10.0到usr/local/cuda然后修改环境变量，然后执行conda install pytorch==1.1.0 torchvision==0.3.0 cudatoolkit=10.0 -c pytorch 才可以

I run the “nvidia-sim” command to check the nvidia support up to cuda 11.2, which should be backward compatible？
Then I have another problem with my cuda. Runfile is downloaded to usr/local/cuda through conda install cudatoolkit=10.0 .and the the commend "nvcc -V" show there is no CUDA. Conda Install mainly contains the dynamically linked libraries that applications rely on to use CUDA-related functionality. The Conda installation only installs some computing libraries, but does not install compilation tools.
so,first, I should Download cuda10.0 to usr/local/cuda and modify the environment variables. the run the commend "conda install PyTorch ==1.1.0 TorchVision ==0.3.0 CUDatoolKit =10.0 -c PyTorch"

[zhengye1995]

Yes, you should install cuda in your system frist and use ncvv -V to check it.

As for as I know, 3090 requires at least cuda11, won't work properly on cuda10.

[wisest-cui]

thanks you for your reply!
and you are right， 3090 requires at least cuda11,I googled it。

[wisest-cui]

I have run the program successfully! thank you for your answer!

[JKZhan]

I have run the program successfully! thank you for your answer!

which version of mmdetection did you install ?