Skip to content
This repository has been archived by the owner on Jul 26, 2019. It is now read-only.

number of trainable parameters #19

Open
andrey999333 opened this issue Feb 2, 2019 · 1 comment
Open

number of trainable parameters #19

andrey999333 opened this issue Feb 2, 2019 · 1 comment

Comments

@andrey999333
Copy link

andrey999333 commented Feb 2, 2019

I don't quite understand one point. When I downloaded your keras representation of BERT and check the number of trainable parameters in summary, it showed ~177 mil parameters, while in official bert it should be 110 mil for base model. Could you explain where this difference comes from?

@Separius
Copy link
Owner

Separius commented Feb 2, 2019

Hi,
I'm not entirely sure, but maybe it's because of the subword embeddings?
most of the time people don't count input embeddings in their model parameters.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants