BP-Transformer: Modelling Long-Range Context via Binary Partitioning

Benchmark Model Rank Results
language-modelling-on-enwiki8BP-Transformer (12 layers)#20Bit per Character (BPC): 1.02Number of params: 38M
language-modelling-on-text8BP-Transformer - 12 Layers#9Bit per Character (BPC): 1.11
sentiment-analysis-on-imdbBP-Transformer + GloVe#31Accuracy: 92.12
sentiment-analysis-on-sst-5-fine-grainedBP-Transformer + GloVe#13Accuracy: 52.71