My question is similar to this previous question from 2013.
What I would like to have is a USB stick with a Linux installation on it such that I can boot from the stick and run PyCUDA programs on the GPU. That is, I need a live Linux stick with CUDA installation.
Is that possible? Has anybody done it? Can it be done such that the CUDA installation remains and keeps working when plugging the stick into another computer?
toram
option on the kernel line that loads so that it loads the entire OS into RAM instead. – Terrance Jan 09 '18 at 20:33