custom_matmul_kernels icon indicating copy to clipboard operation
custom_matmul_kernels copied to clipboard

NoneType' object has no attribute 'split

Open wxthu opened this issue 2 years ago • 3 comments

when I run the kernel code in main branch or master branch, the same error occured as following: image

wxthu avatar Mar 04 '22 14:03 wxthu

please make sure you have nvcc installed on your system, run:

which nvcc

on linuxs, or:

where nvcc

on windows. it should return the path of nvcc.

if nvcc isn't installed on your system, please go to https://developer.nvidia.com/cuda-downloads to download cuda toolkit. Do NOT download v11.6, it isn't supported by cupy yet.

Install the latest version of cupy that matches your cuda toolkit version (or upgrade cupy to the latest version if you have already installed). For example, if your cuda toolkit version is 11.5, run:

pip3 install cupy-cuda115

to install cupy. If your cuda toolkit version is 11.5, but you've already installed cupy-cuda112, first uninstal cupy-cuda-112, then install cupy-cuda115.

if you're confident that nvcc is installed, but the above command doesn't return the path of nvcc, then make sure /usr/local/cuda-{version}/bin (on linux) or C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v{version}\bin (on windows) is included in PATH variable.

DeMoriarty avatar Mar 04 '22 14:03 DeMoriarty

I am sure that I have installed nvcc and it has been included in PATH variable. However, it seems still no work image image

wxthu avatar Mar 05 '22 01:03 wxthu

if /home/oushou/cuda-11.0/ is where cuda is installed, then you should use the nvcc in /home/oushou/cuda-11.0/bin/. also make sure the installed version of cupy is cuda-cupy110

DeMoriarty avatar Mar 05 '22 10:03 DeMoriarty