required version to compile cuda code

Open charliewang789 opened this issue 5 years ago • 3 comments

can someone post a workable version of pytorch, cuda, c++ that can compile the code?

Nov 12 '20 18:11 charliewang789

Worked for me on pytorch 1.4.0. I tested it on google colab.

Feb 23 '21 20:02 GlockPL

How did you make it work @GlockPL ??? I tried installing pytorch 1.4.0 in colab and I have the following error

CUDA error at adaptive_gridsampler_kernel.cu:76 code=209(cudaErrorNoKernelImageForDevice) "cudaGetLastError()"

Mar 03 '21 10:03 davertor

This is my code from colab: !pip install torch==1.4.0 torchvision==0.5.0 !git clone https://github.com/sunwj/CAR.git %cd ./CAR/ !ls %cd adaptive_gridsampler/ !ls !python3 setup.py build_ext --inplace Last part is crucial because you need to compile this adaptive_gridsampler which is the reason why you get this error.

Mar 03 '21 14:03 GlockPL