4 ms·
But for deployment in smaller devices you can use techniques such as distillation, quantization and sparsity. Training and inference are very different problems
by david-gpu 6y ago
But for deployment in smaller devices you can use techniques such as distillation, quantization and sparsity. Training and inference are very different problems in practice.
- dontreact 6y agoYes but you can do that with efficientnet as well. The point is that this is an improvement only for training because it uses computations which are highly optimized on TPU