I’ve been struggling for the past two days to install XGBoost with GPU support for R. I’ve exhausted almost everything I could think of or find on here, the official docs, and on stack overflow.
I’ll try to be as comprehensive as possible!
GPU: Nvidia RTX 2080
Environment (done on fresh install of Windows 10)
Operating System: Windows 10 Pro
Visual studio: Visual Studio 2017 (also tried Visual Studio 2015)
Cmake: cmake 3.13.1 (tried several older versions of cmake as well)
R: R 3.5.1 (also tried 3.3.1, 3.4.3, and a patched 3.4.3)
- I made sure the correct path was set in the environment settings
- I made sure the correct paths were set to the Rtools bin and the mingw64 bin
CUDA: Downloaded the newest CUDA Toolkit
Steps to producing error
The following steps all complete without error:
git clone --recursive https://github.com/dmlc/xgboost
git submodule init
git submodule update
cmake … -DUSE_CUDA=ON -DUSE_NCCL=OFF -DR_LIB=ON
I have also used the following variations without luck before building:
cmake … -G"Visual Studio 15 2017 Win64" -DUSE_CUDA=ON -DR_LIB=ON
cmake … -G"Visual Studio 14 2015 Win64" -DUSE_CUDA=ON -DR_LIB=ON
When I enter the following:
cmake --build . --target install --config Release
I get the following error message:
“C:\Users\John Kilbride\xgboost\build\install.vcxproj” (default target) (1) ->
“C:\Users\John Kilbride\xgboost\build\ALL_BUILD.vcxproj” (default target) (3) ->
“C:\Users\John Kilbride\xgboost\build\gpuxgboost.vcxproj” (default target) (5) ->
(CustomBuild target) ->
C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\IDE\VC\VCTargets\Microsoft.CppCommon.targets(20
9,5): error MSB6006: “cmd.exe” exited with code 1. [C:\Users\John Kilbride\xgboost\build\gpuxgboost.vcxproj]
563 Warning(s) 1 Error(s)
Time Elapsed 00:00:23.24
The text for the warnings is massive. If that will be useful I can create a link to a pastebin/host a .txt somewhere!
I’ve run out of ideas for what to do. I’ve tried using different versions of the software to match the docs as closely as possible w/out luck. I’ve also tried a TON of variations based on what other people had luck using.
Other instances of people talking about this issue concluded with them downloading the unofficial pre-compiled versions which doesn’t exist to my knowledge for R.