Pushing the bounds of dropout

Benchmark Model Rank Results
language-modelling-on-penn-treebank-word2-layer skip-LSTM + dropout tuning#23Test perplexity: 55.3Validation perplexity: 57.1Params: 24M