Build error

[EnricoBeltramo]

I tried to compile under win10, Vs2019, but no way to work.
I have always follow error:
error: command 'C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.26.28801\bin\HostX86\x64\cl.exe' failed with exit status 2

withc version of compiler needs to use for a correct build?

[byq-luo]

I tried to compile under win10, Vs2015, but no way to work.
error: command 'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1

[alvinhu]

I tried to compile under win10, Vs2019, but no way to work.
I have always follow error:
error: command 'C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.26.28801\bin\HostX86\x64\cl.exe' failed with exit status 2
withc version of compiler needs to use for a correct build?

I also got the same error.

[EnricoBeltramo]

To install it under windows I did:

• created a virtual env conda with python 3.7 with follow versions of torch:
  conda install pytorch==1.4.0 torchvision==0.5.0 cudatoolkit=10.1 -c pytorch
• I installed pycocotools from here:
  pip install git+https://github.com/philferriere/cocoapi.git#egg=pycocotools^&subdirectory=PythonAPI
• removed in setup the dipendences from pycoocotools (otherwise try to reinstall if)
• installed visual studio C++ build tools v1.40
• copied nms_rotated_cuda.cu as version for windows (should be used as full replacement, but is not aligned as actual version):
  https://github.com/conansherry/detectron2
• make changes in torch files as specified in issue:
  Can I install detectron2 on Win10? facebookresearch/detectron2#9
  after that, I correctly build and installed (from local copy)

[Tomiha]

I tried to compile under win10, Vs2019, but no way to work.
I have always follow error:
error: command 'C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.26.28801\bin\HostX86\x64\cl.exe' failed with exit status 2
withc version of compiler needs to use for a correct build?

I also got the same error.

Got it working eventually?
Been trying for several hours to get detectron2 to compile but there always are a bunch of errors. Already tried 5 different compilers... Any idea on the exact compiler and dependency versions?

[Tomiha]

To install it under windows I did:

* created a virtual env conda with python 3.7 with follow versions of torch:
  conda install pytorch==1.4.0 torchvision==0.5.0 cudatoolkit=10.1 -c pytorch

* I installed pycocotools from here:
  pip install git+[https://github.com/philferriere/cocoapi.git#egg=pycocotools^&subdirectory=PythonAPI](https://github.com/philferriere/cocoapi.git#egg=pycocotools%5E&subdirectory=PythonAPI)

* removed in setup the dipendences from pycoocotools (otherwise try to reinstall if)

* installed visual studio C++ build tools v1.40

* copied nms_rotated_cuda.cu as version for windows (should be used as full replacement, but is not aligned as actual version):
  https://github.com/conansherry/detectron2

* make changes in torch files as specified in issue:
  [facebookresearch/detectron2#9](https://github.com/facebookresearch/detectron2/issues/9)
  after that, I correctly build and installed (from local copy)

Hey,

what exactly do you mean with 'visual studio C++ build tools v1.40', are you referring to MSVC v140 - VS2015 Build Tools (v14.0) ?