BBuf commented on pull request #8790: URL: https://github.com/apache/tvm/pull/8790#issuecomment-1011613582
> @BBuf so the image `tlcpackstaging/ci_gpu:20211219-100400-c0d326dbd` https://hub.docker.com/layers/tlcpackstaging/ci_gpu/20220112-000102-d1ee20191/images/sha256-6c9bd4cef78df644fa1cee7dd8b2915488572543e2407f29044e38861616b350?context=explore has oneflow installed, but when I test the installation locally I got this error: > > ``` > >>> import oneflow > libibverbs not available, ibv_fork_init skipped > W0112 20:58:52.579394 22 cuda_device_descriptor_class.cpp:48] CUDA driver version is insufficient for CUDA runtime version > >>> oneflow.__version__ > '0.6.0+cu110' > ``` > > Can you test this locally if this is a real problem? The cuda version on CI is > > ``` > # nvcc --version > nvcc: NVIDIA (R) Cuda compiler driver > Copyright (c) 2005-2020 NVIDIA Corporation > Built on Wed_Jul_22_19:09:09_PDT_2020 > Cuda compilation tools, release 11.0, V11.0.221 > Build cuda_11.0_bu.TC445_37.28845127_0 > ```  As you can see from here, installing the CUDA10 version of OneFlow requires the system's CUDA Driver Version>=450.36.06. The CUDA Driver Version of my machine is 460.91.03, which can install and use OneFlow normally. So, what is the CUDA Driver Version of docker image `tlcpackstaging/ci_gpu:20211219-100400-c0d326dbd` ? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
