10 ms·
That’s crazy, I’ve never seen one that dropped whole layers from a pre-trained model. I guess that avoids the sparse matrix math.
by sp332 3y ago
That’s crazy, I’ve never seen one that dropped whole layers from a pre-trained model. I guess that avoids the sparse matrix math.