Closed ivanycwang closed 1 year ago
Hi, can you share your pip list & conda env? Also with your full error messages
OK, my question is that I cannot find the proper version for both torch and torch-scatter well. That is, if my torch works, then torch-scatter cannot work. Otherwise, torch-scatter can work but my torch become unavailable. . The error message for torch_scatter: "ModuleNotFoundError: No module named 'torch_scatter'" I try to use “conda install -c pytorch torch-scatter” to install torch_scatter. . After torch_scatter is installed. I encounter another error message “OSError: libc10_cuda.so: cannot open shared object file: No such file or directory”. . Thank you for answering my issues!
BTW, could you provide the .yml file so I can re-implement your experiment as well? I really need your identical experimental configuration...Thanks!!!
OK, my question is that I cannot find the proper version for both torch and torch-scatter well. That is, if my torch works, then torch-scatter cannot work. Otherwise, torch-scatter can work but my torch become unavailable. . The error message for torch_scatter: "ModuleNotFoundError: No module named 'torch_scatter'" I try to use “conda install -c pytorch torch-scatter” to install torch_scatter. . After torch_scatter is installed. I encounter another error message “OSError: libc10_cuda.so: cannot open shared object file: No such file or directory”. . Thank you for answering my issues!
It seems to be the issue related to torch_scatter
. Maybe you can search for solution in pytorch_scatter.
BTW, you can safely comment out from torch_scatter import scatter
if you dont need FastMDCT4
(and anything related to). The implementation of MDCT
we used in this work is identical with the Matlab's, which is fast enough.
BTW, could you provide the .yml file so I can re-implement your experiment as well? I really need your identical experimental configuration...Thanks!!!
We dont use yml
for configuration. For our experiment configuration, please check ./train.sh
and option descriptions in ./options
. You may just prepare you dataset & indexing files and run the script to train the network.
As mentioned, I have tried uninstalling and reinstalling torchaudio, but I still encounter the same issue. I have been stuck with this problem for a while, and I would be very grateful if someone could help me resolve it. The system I am using is Ubuntu with CUDA 12.0.