Fraternal Dropout

Benchmark Model Rank Results
language-modelling-on-penn-treebank-wordAWD-LSTM 3-layer with Fraternal dropout#27Test perplexity: 56.8Validation perplexity: 58.9Params: 24M
language-modelling-on-wikitext-2AWD-LSTM 3-layer with Fraternal dropout#28Test perplexity: 64.1Validation perplexity: 66.8Number of params: 34M