Machine Translation on MultiUN (test)
50.8X->En BLEUBranchNorm
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| BranchNorm#Layers=1000, #Params=3.8B2023.05 | 50.8 | 57.6 | 54.2 | |
| DeepNorm (reimplemented)#Layers=1000, #Params=3.8B2023.05 | 50.3 | 57.2 | 53.8 | |
| Pre-LN#Layers=1000, #Params=3.8B2023.05 | 50.1 | 56.7 | 53.4 | |
| BranchNorm#Layers=200, #Params=863M2023.05 | 49.7 | 57.2 | 53.4 | |
| DeepNorm (reimplemented)#Layers=200, #Params=863M2023.05 | 49.2 | 56.9 | 53.1 | |
| Pre-LN#Layers=200, #Params=863M2023.05 | 49.1 | 56.3 | 52.7 | |
| Baseline#Layers=24, #Params=173M2023.05 | 46.1 | 53.9 | 50 | |
| Baseline#Layers=12, #Params=133M2023.05 | 43.8 | 52.3 | 48.1 |