What is the CUDA version of the Ubuntu System during the testing period?

[ZAX130]

My model uses a custom cuda operator, so the cuda version in pytorch needs to be consistent with the cuda version of the system at runtime. I would like to know what the cuda version of the system is.

[s-sd]

@ZAX130 we use the following versions:

('cuda_version', '10.1'), ('cudnn_version', '7')

However, since we accept docker containers for submissions, feel free to use whatever version you want as long as the docker container is able to run the testing script which uses tensorflow==2.4.1.

Alternatively, leave us an instructions.txt telling us the exact versions or install instructions for your env and we'll try to accommodate however we can. Hope that helps. :)