Machine Translation on zh-en stylized (test)
3.6Modern (M)Baseline
Evaluation results
| Method | Links | ||
|---|---|---|---|
| BaselineArchitecture=Transformer-base, Encoder layers=6, Decoder layers=6, Attention layer size=512, Batch size=4096 tokens, Optimizer=Adam, Label smoothing=0.12022.12 | 3.6 | 3.7 | |
| StyleAPArchitecture=Transformer-base, Encoder layers=6, Decoder layers=6, Attention layer size=512, Batch size=4096 tokens, Optimizer=Adam, Label smoothing=0.12022.12 | 3.5 | 3.6 |