Code largely from: https://github.com/deeplearning4j/dl4j-examples
Goal: To experiment tuning/using DL4J. To enable this, have largely taken code samples from DL4J repo, with the aim to tune/build upon.
https://deeplearning4j.org/lstm.html http://karpathy.github.io/2015/05/21/rnn-effectiveness/
-Add benchmarking in to track how long it takes to train. -Load params from JSON. -Try optimizing methods as per guide. -Speed up training - optimize JVM.
Tried: Changing JVM options (fill in) - no changes. Approx run time ~ 45 mins. (1 epoch)
Default DL4J settings: Approx run time ~ 45 mins. (1 epoch) Set in pom.xml: <nd4j.backend>nd4j-native-platform</nd4j.backend>
Changing to CUDA: Approx run time ~ 15 mins. (1 epoch) Set in pom.xml: <nd4j.backend>nd4j-cuda-8.0-platform</nd4j.backend> https://developer.nvidia.com/cuda-downloads