search

allenai / bilm-tf

Tensorflow implementation of contextualized word representations from bi-directional language models
Apache License 2.0
1.62k stars 452 forks source link

[Question] Perplexity at the end of training #196

Closed mmarius closed 5 years ago

mmarius commented 5 years ago

In the Deep contextualized word representations paper you report an average forward and backward perplexity of 39.7. Is this on the validation data or on the training data?

matt-peters commented 5 years ago

Validation

mmarius commented 5 years ago

Thanks!