search

rsennrich / subword-nmt

Unsupervised Word Segmentation for Neural Machine Translation and Text Generation
MIT License
2.18k stars 464 forks source link

Add note about BPE-Dropout during training #100

Closed VProv closed 3 years ago

VProv commented 3 years ago

Dear Rico,

I have been approached several times by people who have learned models by applying BPE-Dropout only once on the training corpus. I think this note can save many GPU-hours in the future.

Ivan

rsennrich commented 3 years ago

thanks!