Pardon my lack of knowledge here. I am really new to the ML/DL world, and decided to jump in and have a go. Not really helped by my lack of expertise with Linux… All that aside though.
My end goal is to get Resnet50 running, I want to test training times between multiple GPUs. Problem is, I can’t get the nvcr caffe2 image to recognize Turing GPUs. Attempting on my 2080ti (haven’t tried a different GPU yet), and every time I start the image it tells me No Supported GPUs detected to run this container.
In the base OS, the driver is installed just fine and so far everything is working. This is a fresh install of Ubuntu 18.08, 415 drivers, and latest kernel. SMI within the container (guess this is called a journal?) is showing the GPU, but thats about it.
Anyone found a solution to this yet?