3 ms·
Does it have "fake" tensor cores? Aren't those for training?
by wmf 7y ago
Does it have "fake" tensor cores? Aren't those for training?
- fizixer 7y agoYou still need tensor cores for inference. But they don't do weight updates. Learning/training is all about updating the weights (through backpropagation or whatever). So another way to put it: its tensor cores do feed-forward calculations, but no backpropagation, and no weight updates.
- rrss 7y agoThe hardware and platform is capable of training just fine. It's just rarely done because it is slower than training on pretty much any discrete GPU.