I know that installing FFCV on CPU-only node is not officially supported at this point.
However, I'm wondering there is anything in FFCV that would conceptually prevent us using FFCV on CPU-only nodes?
(e.g. some optimization/functionality code that is deeply coupled with CUDA/GPU to compile).
(e.g. does FFCV offload some data preprocessing on GPU? If so, the compilation of that part of code might potentially be a problem.)
Or do you think it is conceptually doable, but would take time to figure out the correct environments?
Hi,
I know that installing FFCV on CPU-only node is not officially supported at this point.
However, I'm wondering there is anything in FFCV that would conceptually prevent us using FFCV on CPU-only nodes? (e.g. some optimization/functionality code that is deeply coupled with CUDA/GPU to compile). (e.g. does FFCV offload some data preprocessing on GPU? If so, the compilation of that part of code might potentially be a problem.)
Or do you think it is conceptually doable, but would take time to figure out the correct environments?
Thanks! Meng