Machine Translation on WAT2017 Small-NMT English-Japanese (test)
43.68BLEUTransformer
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TransformerDecoding strategy=beam-search (beam=4), Model architecture=Transformer base2019.05 | 43.68 | 306 | 22.6 | |
| Levenshtein TransformerTraining teacher=distillation (autoregressive teacher model), Model architecture=Transformer base2019.05 | 43.17 | 106 | 1.97 | |
| TransformerDecoding strategy=greedy, Model architecture=Transformer base2019.05 | 42.86 | 261 | 22.6 | |
| Levenshtein TransformerTraining teacher=oracle, Model architecture=Transformer base2019.05 | 42.36 | 112 | 2.61 |