| 16 Sep 2022 |
 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 |
| 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..?🤔
But in the github history, it seems like there was no problem building the package without cuda_cudart which means that I'm probably doing something wrong or unnecessary.
| 18:13:12 |
| 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
)
The above code cause error 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
# 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/
I mentioned. Am I right..?🤔
But in the github history, it seems like there was no problem building the package without `cuda_cudart` which means that I'm probably doing something wrong or unnecessary.
| 18:14:59 |
| 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
)
The above code cause error 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
# 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/
I mentioned. Am I right..?🤔
But in the github history, it seems like there was no problem building the package without `cuda_cudart` which means that I'm probably doing something wrong or unnecessary.
| 18:15:22 |
| 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
)
The above code cause error 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
# 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/
I mentioned. Am I right..?🤔
But in the github history, it seems like there was no problem building the package without `cuda_cudart` which means that I'm probably doing something wrong or unnecessary.
| 18:15:52 |
| 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
)
The above code cause error 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 following code
# 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/
I just mentioned. Am I right..?🤔
But in the github history, it seems like there was no problem building the package without cuda_cudart which means that I'm probably doing something wrong or unnecessary.
| 18:16:20 |
| 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
)
The above code block causes error 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 following code
# 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/
I just mentioned. Am I right..?🤔
But in the github history, it seems like there was no problem building the package without cuda_cudart which means that I'm probably doing something wrong or unnecessary.
| 18:21:50 |
 SomeoneSerge (back on matrix) | pytorch derivation uses symlinkJoin which includes contents of cudatoolkit.out and cudatoolkit.lib | 20:54:46 |
| Suwon Park | Someone S: Aha! Let me try some modification! Thank you! | 21:10:06 |
| Suwon Park | * Someone S: Aha! Let me try some modifications! Thank you! | 21:10:18 |
| SomeoneSerge (back on matrix) | cudatoolkit.{out,lib} bring in a lot (4-5GiB) of luggage; if you'd like to get rid of it later, maybe you could start with https://github.com/NixOS/nixpkgs/blob/befe56a1ee1d383fafaf9db41e3f4fc506578da1/pkgs/development/python-modules/pytorch/default.nix#L57 | 21:14:47 |
| SomeoneSerge (back on matrix) | * cudatoolkit.{out,lib} brings in a lot (4-5GiB) of luggage; if you'd like to get rid of it later, maybe you could start with https://github.com/NixOS/nixpkgs/blob/befe56a1ee1d383fafaf9db41e3f4fc506578da1/pkgs/development/python-modules/pytorch/default.nix#L57 | 21:14:53 |
| 17 Sep 2022 |
 aidalgol | In a flake shell with config.allowUnfree = true; and config.cudaSupport = true;, the python torch module is throwing an unknown CUDA error. Is there something more I need to do to get the package's CUDA support enabled?
File "/nix/store/bf48f3zny7q08lg4hc4279fn3jw1lkpl-python3-3.10.6-env/lib/python3.10/site-packages/torch/cuda/__init__.py", line 217, in _lazy_init
torch._C._cuda_init()
RuntimeError: CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero.
| 05:54:58 |
 tpw_rules | are you running on nixos? | 05:55:22 |
| aidalgol | Yes, sorry, this is on NixOS. | 05:55:35 |
| tpw_rules | and you have the nvidia drivers set up and nvidia-smi works and stuff? | 05:57:16 |
| aidalgol | Yep, nvidia-smi output still looks good. | 05:58:14 |
| tpw_rules | is this the torch-bin module or did you compile it yourself? | 05:58:52 |
| aidalgol | I was referencing torch, not torch-bin. Should I try that one? | 06:00:26 |
| aidalgol | I'm also using the cuda-maintainers cachix cache, if that makes a difference. | 06:01:19 |
| tpw_rules | you can try torch-bin, it's precompiled by upstream with cuda support. | 06:01:41 |
| tpw_rules | do you have an excessively recent nvidia card? | 06:02:01 |
| aidalgol | RTX3080 | 06:02:07 |
