In the spirit of constructive criticism, I'd suggest that the GPU section of the summary feels like a fish out of water and should probably either be removed or supplemented with a bit more context from the actual paper. Cheers.