List your company

Masked Language Modeling on BERT Pretraining Corpus

2.89PerplexitygMLP_xlarge
2.83443.20973.5853.9603May 17, 2021

Evaluation results

MethodLinks
gMLP_xlargeParams (M)=9412021.052.89
aMLP_largeAttn Size=128, Params (M)=3162021.053.19
gMLP_largeParams (M)=3652021.053.32
BERT_large (ours)Attn Size=1024 (64 x 16), Params (M)=3362021.053.35
aMLP_baseAttn Size=64, Params (M)=1092021.053.95
BERT_base (ours)Attn Size=768 (64 x 12), Params (M)=1102021.054.17
gMLP_baseParams (M)=1302021.054.28