-
Notifications
You must be signed in to change notification settings - Fork 8k
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
windows; CUDA Compiler not found #2971
Comments
Hi, it compiles - but without cuda support. So; This is a fresh W10 Client. i did setup VS2019 community edition; Downloaded Cuda, opencv and all with vcpk; Added a ton of environment variables; So far it compiles to a running darknet.exe - but without cuda support. It does find cuda - but not the cuda compiler. i did find the tipp to uninstall VS and cuda - and did so - but it does not have an impact in the not finding of cmake of the nvcc file. which is in $path... so im stuck and with no idea left and would appriciate any help many times.
|
@JoeRu Hi,
|
Hi @AlexeyAB
... |
hmm. thats from cmake documentation; Even this warning is wrong
|
CUDA nsight plugin isn't installed in MSVS. Just re-install CUDA. |
There was a bug inside vcpkg for CUDA. I fixed it not long ago, please do a git pull and then ./bootstrap-vcpkg.ps1 again to rebuild vcpkg with the bugfix. This is in case you are building darknet through vcpkg |
A docker container and distribution, for Linux CPU, Linux GPU and Windows (only CPU unfortunately, gpu is unsupported in native windows containers as of now) will come soon, I hope. It depends on how much busy I will be (probably a lot), if I have to do it alone |
@JoeRu Sorry for the chat, I am on phone. Please let me know if rebuilding vcpkg solved your problems (it should 😉) |
@JoeRu up until one week ago, it was impossible to build darknet with Cuda inside vcpkg, due to a bug in vcpkg. Now it is fixed and you have to update your vcpkg installation (pulling and rebuilding it) to fix it. Building darknet inside vcpkg means doing just
It downloads darknet from this repository, setup dependencies automatically (stb, pthreads, OpenCV and CUDA), and builds everything, in a single step.
Let's understand what's going on: the first two lines come from darknet, which uses the modern At the beginning I was thinking, because of the symptoms, that you were building darknet inside vcpkg, but now I realised that you're not, since you are describing also the output of Unfortunately, I have to say that Visual Studio 2019 + CUDA 10.1 is still unsupported. I had the opportunity to test and verify Visual Studio 2019, but without CUDA. |
OK; So - some experiments on my side - with the following results. i did an uninstall of VS2019 and did an installation of VS2017; i did a pull of https://github.com/Microsoft/vcpkg with the master branch;
But only CPU support - no CUDA. So if this would compile with CUDA this would be a very nice and easy way to install darknet NICE On the behaviour of the build1.ps1 did nothing change. I did lost track of all the environment variables - so there is actual the following set |
so - solved. For future references and to check: Dependency : VS 2017 - vs 2019 is not working (yet) if CUDA doesn't work For more detailed checking :
|
good that it's working now. |
I still don't understand what happened in your installation. If you want to check, in continuous integration we verify for each commit building on linux, macOS, Windows, with and without CUDA, and it is properly working in all configurations. I will soon add checks for CUDA and VS2019, so to verify progress with this configuration |
hi @cenit i will do another fresh pull of VCPKG and try it to install the suggested way - with all environment removed. But it takes a lot of time to compile - so i will report on later point on time. |
don't need to rebuild all |
Just to close this one and maybe to update Documentation for Windows: Download Visual-Studio 2017 and install;
Add to environment variable PATH the following: Now you should be able to start darknet.exe from every fresh CMD or Powershell. |
exactly what I meant and what we aim in a not-so-distant future. Maybe automatically done by a script 😄 (it's still a little bit experimental to declare it ready 😉) |
and to update, you just need to go to |
I'm having similar issue (#3028). |
My environment: Win10 + cuda 10.1 + cuddn + Visual Studio 2019. I think I found the reason, this is problem between the cuda and Visual Studio. Because the cuda should be installed after the Visual Studio. When cuda is installed there will be a item for finding whether you have installed Visual Studio before, if you did then cuda adapting the environment, if you didn't you can ignore. Like #2971 (comment) You can reinstall the cuda to solve the problem. There is a way to test if cuda is satisfied to Visual Studio:
you will find a item named The cmake will like this: |
No description provided.
The text was updated successfully, but these errors were encountered: