3 ms·
Has anyone actually managed to get these to work on an imagenet scale?
by deepnotderp 10y ago
Has anyone actually managed to get these to work on an imagenet scale?
- ar15saveslives 10y agoI don't think it's usable IRL (in its present state), as, according to figures, it doesn't work well even on cifar/mnist. Correct me if I'm wrong, but the value of this paper is that you can decouple a model and train layers asynchronously/independently, just first steps to distributed NN training.
- curuinor 10y agoWell, HogWild works well enough. It's another one for the giant "weird shit you can get away with in neural network land" bucket, I think
- williamtrask 10y agoin the paper, it seemed to work quite well for LSTMS (language modeling), although it is best suited for tasks with very, very large neural networks for which parallelism is beneficial