-
Notifications
You must be signed in to change notification settings - Fork 1.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
pybind.obj : error LNK2001: 无法解析的外部符号 "class at::Tensor __cdecl nms_rotated(class at::Tensor,c lass at::Tensor,class at::Tensor,class at::Tensor,float,int)" (?nms_rotated@@YA?AVTensor@at@@V12@V12@11MH@Z) #709
Comments
Hello, Please see if this modification solves your problem. It is based on version v1.2.2 . |
I have tried but unfortunately the error still happened:
|
Try remove the build folder (D:\Github\mmcv\build) and rebuild? |
Also, as a workaround, if m.def("nms_rotated", &nms_rotated, "NMS for rotated boxes", py::arg("dets"),
py::arg("scores"), py::arg("order"), py::arg("dets_sorted"),
py::arg("iou_threshold"), py::arg("multi_label")); But I think the problem lies here. |
Thank you! I'll try later. Does it affect mmdetection training? |
I think no. Moreover, I upload the compiled wheel package here (miniconda with python 3.8, torch 1.7, cuda 11.0.3). You can also try pip install this package directly. |
Thank you very much! I have successfully installed
Next time I will try to train my own dataset to see whether it work properly. |
In my computer I have successfully trained my own datasets. But when I configured in another computer (CUDA10.2, Win10, Python3.8, pytorch1.7.1), the error still occured. I tried some ways:
m.def("nms_rotated", &nms_rotated, "NMS for rotated boxes", py::arg("dets"),
py::arg("scores"), py::arg("order"), py::arg("dets_sorted"),
py::arg("iou_threshold"), py::arg("multi_label")); |
I reinstalled cuda11 and pytorch1.7.1+cu110, now I have accounted another problem when I excute the training script:
|
What GPU do you use? I compiled cuda code for sm61 (GTX 10xx GPUs) target only. The package can be compiled to support all GPUs but only for specific vertion of python, pytorch and cuda. So if you change either of them you need a recompilation the package. Moreover, the modification I mentioned in the previous reply means that you need to first clone my fork and then checkout to the fixed branch. The master version is NOT fixed. Which meas you need to first
and then
and then launch compilation. Did you remember to checkout to correct branch? |
Both of the two computers' GPU are RTX 2080. In my computer it works, but the other does not (same CUDA, VS2019, pytorch, python). I think the problem is mainly in |
My mistake! I haven't changed the branch before. It can be compiled using the branch |
I have the same error during the installation:
here is the environment info:
on Windows 10, vs2019. MSVC\14.29.30133 |
fixed by #1524 |
same problem need help |
I have same problem, while my GPU is RTX4060 |
OS: Win10
VS: 2019 (19.28.29335)
mmcv: 1.2.1
CUDA: 11.0
pytorch: 1.7.1
Python: 3.8.5 (tags/v3.8.5:580fbb0, Jul 20 2020, 15:57:54) [MSC v.1924 64 bit (AMD64)]
本人编译mmcv-full的环境介绍如上,但是出现了以下错误:
看提示好像是link.exe的问题,奇怪的是编译的时候host是x86,不知道是不是这个问题,但是试了一下没有找到build_ext时手动设置msvc的host的方法,请问有人在Windows下成功编译了mmcv-full了吗?mmdetection依赖这个,否则用不了
The text was updated successfully, but these errors were encountered: