I am using a NVIDIA DGX machine that has Cuda 9 installed in it.
I am setting the variables PATH, NUMBAPRO_NVVM , NUMBAPRO_LIBDEVICE and LD_LIBRARY_PATH
When I try to run my code, I am getting the following error. When Cuda 9 is present, why I am getting an error to install Cuda 7.5 ?
RuntimeError: Missing libdevice file for compute_50.
Please ensure you have package cudatoolkit 7.5.
Install package by:
conda install cudatoolkit=7.5
When I look in nvvm/libdevice directory, there is only one file (libdevice.10.bc) . There is no file for compute_50
because your numba installation is linked against CUDA 7.5
I don’t recommend attempting to install CUDA 7.5 on your DGX.
Instead, you should find an alternate tool chain install method for Pycuda that is compatible with CUDA 9. If you install pycuda by installing binaries, those binaries are linked against CUDA 7.5
Instead you would need to install/build pycuda from sources, or find an install method for pycuda that installs binaries linked against CUDA 9.
@txbob , Thanks.
I downloaded the Pycuda manually and installed it and now I am getting a new error
numba.cuda.cudadrv.error.CudaSupportError: Error at driver init:
CUDA driver library cannot be found.
If you are sure that a CUDA driver is installed,
try setting environment variable NUMBAPRO_CUDA_DRIVER
with the file path of the CUDA driver shared library.
Is there a way to fix this