guyue huang

Issue #7

It seems that 'cmake ' cannot go through, so 'make' fails. Try the following steps ``` git clone --recursive https://github.com/hgyuhungry/ge-spmm # this step may give error about gunlock's submodule, but...

It should work together with cudatoolkit v10.1. Try this inside your Python environment `conda install pytorch==1.4.0 torchvision==0.5.0 cudatoolkit=10.1 -c pytorch`

Hi @GugaGugaGuga, the error occurs because cusparse in cuda11 and cuda10 have different APIs, so what matters is CUDA Toolkit version. Can you try `print(torch.version.cuda) ` in python and see...

Sorry, the `torch.version.cuda` does not matter. The compilation of the shared library spmm.so is through your system's default cuda. If your default nvcc is >= 11, there would be a...

> > Sorry, the `torch.version.cuda` does not matter. The compilation of the shared library spmm.so is through your system's default cuda. If your default nvcc is >= 11, there would...

> ``` > export PATH="/usr/local/cuda-10.1/bin:$PATH" > export LD_LIBRARY_PATH="/usr/local/cuda-10.1/lib64:$LD_LIBRARY_PATH" > export CUDA_HOME="/usr/local/cuda-10.1" > ``` > > > > > > Yes Since I cannot reproduce the environment problem, I suggest you...

> > When I run `"python3.8 gcn_custom_2layer.py --n-hidden=32"`, the following situation occurred: > > ``` > > Using /tmp/torch_extensions as PyTorch extensions root... > > Detected CUDA files, patching ldflags...

The script does not have non-deterministic code, and observations should not be different run-to-run. What you observed possibly signals a GPU memory leak, but I verified that this script causes...

Hi The following figure explains how this kernel arranges thread Idx dimension and block Idx dimension.  If (blockIdx.y