Is UFF file portable across different GPU?

A simple question here:

As we know, the final generated plan file for tensorRT is not portable across different GPU, or if a tensorRT plain file/model is generated in GPU of type A, it will not work in GPU of type B.
But how about the UFF file? Do we need to generate UFF file for each GPU?? OR it’s portable across different GPUs??

BEst
zhen

Hi @xiamenhi,

Yes, the UFF file should be portable across different GPUs and Operating Systems (Windows <-> Linux).