2 ms·
By looking at the paper, memory needed per layer seems to be higher than transformer architecture. Pretty sure that would be blowing up the vram of gpu at scale
by darkbatman 1y ago
By looking at the paper, memory needed per layer seems to be higher than transformer architecture. Pretty sure that would be blowing up the vram of gpu at scale.