| 15 Sep 2022 |
|  m_algery left the room. | 12:33:43 |
 hexa | https://github.com/NixOS/nixpkgs/pull/191333 | 16:22:20 |
 SomeoneSerge (back on matrix) | In reply to @hexa:lossy.network
https://github.com/NixOS/nixpkgs/pull/191333 https://hercules-ci.com/github/SomeoneSerge/nixpkgs-unfree/jobs/1925 | 16:36:11 |
| hexa | fwiw: possibly a bit too early, given that we have a failing numpy build | 16:45:30 |
| hexa | numba pins setuptools<60 | 21:19:53 |
| hexa | i can't even. | 21:20:00 |
| hexa | we have been >60 for a while now | 21:20:08 |
| 16 Sep 2022 |
|  Suwon Park joined the room. | 17:28:32 |
| Suwon Park left the room. | 17:30:03 |
| Suwon Park joined the room. | 17:30:26 |
| Suwon Park | Hello everyone! | 17:33:48 |
| Suwon Park | Is there anyone who has tried building pytorch with cuda enabled? | 17:34:56 |
| Suwon Park | When I tried to build pytorch, this error occurs.
| 17:36:51 |
| Suwon Park | Redacted or Malformed Event | 17:36:59 |
| Suwon Park | * When I tried to build pytorch, this error occurs in the cmake phase.
-- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "11.6")
| 17:37:25 |
| Suwon Park | So I looked up the source of pytorch in nixpkgs and there was no cudaPackages.cuda_cudart in buildInputs. | 17:38:18 |
| Suwon Park | cudaPackages.cuda_cudart seems to fix the error, but is there any reason why it is not there? | 17:39:17 |
| SomeoneSerge (back on matrix) | python3Packages.pytorch currently still uses the older cudaPackages.cudatoolkit expression, which ships a lot of stuff, including cuda_cudart | 17:39:58 |
| SomeoneSerge (back on matrix) | cuda_* expressions are preferred | 17:40:20 |
| SomeoneSerge (back on matrix) | I didn't catch this though. Are you getting this when manually building pytorch, or when running nix build? | 17:41:52 |
| Suwon Park | I'm running nix develop genericBuild! | 17:49:08 |
| Suwon Park | * I'm running nix develop and genericBuild with poetry inside flake.nix! | 17:49:37 |
| Suwon Park | * I'm running nix develop and genericBuild with flake.nix! | 17:49:47 |
| Suwon Park | In reply to @ss:someonex.net
python3Packages.pytorch currently still uses the older cudaPackages.cudatoolkit expression, which ships a lot of stuff, including cuda_cudart But, if you check pytorch in 22.05 version of nixpkgs, the following code already exists
# Move some libraries to the lib output so that programs that
# depend on them don't pull in this entire monstrosity.
mkdir -p $lib/lib
mv -v $out/lib64/libcudart* $lib/lib/
which means that cudaPackages.cudatoolkit expression does not ship with cudart!
| 17:55:09 |
| Suwon Park | In reply to @ss:someonex.net
python3Packages.pytorch currently still uses the older cudaPackages.cudatoolkit expression, which ships a lot of stuff, including cuda_cudart * But, if you check cudaPackages in 22.05 version of nixpkgs, the following code already exists
# Move some libraries to the lib output so that programs that
# depend on them don't pull in this entire monstrosity.
mkdir -p $lib/lib
mv -v $out/lib64/libcudart* $lib/lib/
which means that cudaPackages.cudatoolkit expression does not ship with cudart!
| 17:56:11 |
| Suwon Park | * But, if you check cudaPackages.cudatoolkit in 22.05 version of nixpkgs, the following code already exists
# Move some libraries to the lib output so that programs that
# depend on them don't pull in this entire monstrosity.
mkdir -p $lib/lib
mv -v $out/lib64/libcudart* $lib/lib/
which means that cudaPackages.cudatoolkit expression does not ship with cudart!
| 17:56:20 |
| Suwon Park | If you unpack python39Packages.pytorch (current version : 1.11.0), and go to source/cmake/Modules_CUDA_fix/upstream/FindCUDA.cmake line 1128, there is the following code block which creates -- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "11.6") error.
find_package_handle_standard_args(CUDA
REQUIRED_VARS
CUDA_TOOLKIT_ROOT_DIR
CUDA_NVCC_EXECUTABLE
CUDA_INCLUDE_DIRS
${CUDA_CUDART_LIBRARY_VAR}
VERSION_VAR
CUDA_VERSION
)
And in the end if I understood the code correctly, that ${CUDA_CUDART_LIBRARY_VAR} looks for libcudart.so inside cudaPackages.cudatoolkit which now doesnt have libcuda.*` because of the above code I mentioned. Am I right..?🤔
| 18:07:46 |
| Suwon Park | * If you unpack python39Packages.pytorch (current version : 1.11.0), and go to source/cmake/Modules_CUDA_fix/upstream/FindCUDA.cmake line 1128, there is the following code block which creates -- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "11.6") error.
find_package_handle_standard_args(CUDA
REQUIRED_VARS
CUDA_TOOLKIT_ROOT_DIR
CUDA_NVCC_EXECUTABLE
CUDA_INCLUDE_DIRS
${CUDA_CUDART_LIBRARY_VAR}
VERSION_VAR
CUDA_VERSION
)
That's because in the end, if I understood the code correctly, ${CUDA_CUDART_LIBRARY_VAR} looks for libcudart.so inside cudaPackages.cudatoolkit which now doesnt have libcuda.*` because of the above code I mentioned. Am I right..?🤔
| 18:08:28 |
| Suwon Park | * If you unpack python39Packages.pytorch (current version : 1.11.0), and go to source/cmake/Modules_CUDA_fix/upstream/FindCUDA.cmake line 1128, there is the following code block which creates -- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "11.6") error.
find_package_handle_standard_args(CUDA
REQUIRED_VARS
CUDA_TOOLKIT_ROOT_DIR
CUDA_NVCC_EXECUTABLE
CUDA_INCLUDE_DIRS
${CUDA_CUDART_LIBRARY_VAR}
VERSION_VAR
CUDA_VERSION
)
That's because in the end, if I understood the code correctly, ${CUDA_CUDART_LIBRARY_VAR} looks for libcudart.so inside cudaPackages.cudatoolkit which now doesn``t have libcuda.* because of the above code I mentioned. Am I right..?🤔
| 18:08:53 |
| Suwon Park | * If you unpack python39Packages.pytorch (current version : 1.11.0), and go to source/cmake/Modules_CUDA_fix/upstream/FindCUDA.cmake line 1128, there is the following code block which creates -- Could NOT find CUDA (missing: CUDA_CUDART_LIBRARY) (found version "11.6") error.
find_package_handle_standard_args(CUDA
REQUIRED_VARS
CUDA_TOOLKIT_ROOT_DIR
CUDA_NVCC_EXECUTABLE
CUDA_INCLUDE_DIRS
${CUDA_CUDART_LIBRARY_VAR}
VERSION_VAR
CUDA_VERSION
)
That's because in the end, if I understood the code correctly, ${CUDA_CUDART_LIBRARY_VAR} looks for libcudart.so inside cudaPackages.cudatoolkit which now doesn`t have libcuda.* because of the above code I mentioned. Am I right..?🤔
| 18:09:06 |
