DeLighT

Reported on 5 benchmarks across 2 tasks · 1 paper · 2 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Natural Language Processing4 results

Machine TranslationonWMT2016 English-French
BLEU score· 2020-08-03
40.5
SOTA
DeLighT: Deep and Light-weight Transformer arXiv:2008.00623
Machine TranslationonWMT2016 English-Romanian
BLEU score· 2020-08-03
34.7
SOTA
DeLighT: Deep and Light-weight Transformer arXiv:2008.00623
Machine TranslationonIWSLT2014 German-English
BLEU score· 2020-08-03
35.3
best: 40.43 (PiNMT)
DeLighT: Deep and Light-weight Transformer arXiv:2008.00623
Machine TranslationonWMT2016 English-German
BLEU score· 2020-08-03
28
best: 40.68 (MADL)
DeLighT: Deep and Light-weight Transformer arXiv:2008.00623

Medical1 result

Language ModellingonWikiText-103
Test perplexity· 2020-08-03
24.14
best: 2.4 (RETRO (7.5B))
DeLighT: Deep and Light-weight Transformer arXiv:2008.00623