Neural Machine Translation on IWSLT14 German-to-English small (test)
35.9BLEUPre-Norm Transformer
Evaluation Results
| Method | Links | |
|---|---|---|
| Pre-Norm TransformerArchitecture=Small, Normalization=PN2020.03 | 35.9 | |
| Pre-Norm TransformerArchitecture=Small, Normalization=LN2020.03 | 35.5 | |
| Pre-Norm TransformerArchitecture=Small, Normalization=PN-V2020.03 | 35.5 | |
| Dynamic ConvArchitecture=Small2020.03 | 35.2 | |
| Fixup-InitArchitecture=Small2020.03 | 34.5 | |
| TransformerArchitecture=Small2020.03 | 34.4 | |
| DS-InitArchitecture=Small2020.03 | 34.4 | |
| Pre-Norm TransformerArchitecture=Small, Normalization=BN2020.03 | 34.4 |