As the title says, my CUDA version is 11.7, and the Pytroch version is 1.13.1. Using pytorch’s torch.cuda.is_available() function returns True, but when I use pytorch to prepare to use the GPU, this error is prompted:
dgl._ffi.base.DGLError: [07:40:24] /opt/dgl/src/runtime/c_runtime_api.cc:82: Check failed: allow_missing: Device API cuda is not enabled. Please install the cuda version of dgl.
After I tried to delete dgl (and use import dgl to confirm that the moudle has not been found), I reinstalled the dgl corresponding to the cuda version (ie 11.7) on the official website, but still prompted the above error.
How should I solve this problem? Thanks a lot for your answers!
What was your command to install DGL? The error looks like you installed a CPU version of DGL.
I use the installation command from the official website：
pip install dgl -f https://data.dgl.ai/wheels/cu117/repo.html
could you check whether cuda enabled DGL is really installed? pip3 list|grep dgl
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.