-
Notifications
You must be signed in to change notification settings - Fork 3
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
Failed to build cuvec #47
Comments
Could you try |
Hello! Of course - thanks for replying! Please see below:
|
interesting... based on NVlabs/instant-ngp#119 this looks like a compiler version issue. |
Indeed! I've just purged cuda from my system and reinstalled the runtime (version 12.8) and toolkit (11.5.1) and am still getting the same error (will repeat below just in case there's a subtle difference!). If it's of any help at all, I can get everything working using the docker image that you uploaded on a previous issue. https://github.com/NiftyPET/NIPET/pkgs/container/nipet BW Oli
|
But from the linked issue, I gather trying a different |
I'm using 11, but tried 12 yesterday but that didn't seem to help. Can try again though |
What about |
I downgraded gcc and gpp (just doing gcc led to cplus.1 not being found?).
|
I'm unsure what's wrong with your system... the docker image has: root@niftypet:/# gcc --version
gcc (Ubuntu 11.3.0-1ubuntu1~22.04.1) 11.3.0
root@niftypet:/# nvcc --version
Built on Tue_Jun_13_19:16:58_PDT_2023
Cuda compilation tools, release 12.2, V12.2.91
Build cuda_12.2.r12.2/compiler.32965470_0
root@niftypet:/# python --version
Python 3.10.12 |
OK - I've tried a VE with the same config as that and am still having problems - so it's likely an issue with my environment. I'll try reinstalling ubuntu and see if it works then. |
Based on NVIDIA/nccl#650 perhaps modifying the system headers or forcing GCC 10? pip install -Ccmake.define.CMAKE_C_COMPILER=gcc-10 -Ccmake.define.CMAKE_CXX_COMPILER=g++-10 "git+https://github.com/AMYPAD/CuVec@wrapt" |
OK! Go there. So forcing gcc-10 did not work - however I commented out the suggested lines in usr/include/c++/*/bits/std_function.h (this comment NVIDIA/nccl#650 (comment)). However I could only find these lines in usr/include/c++/11/bits/std_function.h and not /10/bits - so I then removed all traces of cuvec from site packages and ran the following successfully:
Not sure what the long term solution is - but I am now up and running - thanks very much! |
Hi folks,
I've been trying to install AMYPAD and NiftyPET packages - both of which have cuvec as a dependency.
I'm using Ubuntu 22:04LTE with an NVIDIA GTX 1660.
Whenever installing cuvec is attempted I get the following error:
It seems to be failing at importing wrapt? But wrapt is definitely installed on this system (and the VE that it is running in).
I can build cuvec manually using cmake and install using make install:
so it seems to be an issue with the pip installation pipeline.
Can anyone see where I might be going wrong here?
Many thanks!
Oli
The text was updated successfully, but these errors were encountered: