Translation
Transformers
Inference Endpoints

46GB

#2
by Dario9420 - opened

Hi!
Is it normal that when I use XCOMET-XXL the file I get weights more than 46GB?
I was just trying it and wanted to evaluate 1 segment and I got a 46GB file - I used --to_json flag
thanks a lot and sorry if my question is silly!

image.png

Sorry it's not 46!! But it is still pretty heavy... 43GB

Unbabel org

Hey @Dario9420 , yes its normal. The model uses a 10.7B encoder... its a pretty big model

Hi, Ricardo, thank you for your quick response!
Any advice to open the file?
When I try to open it with notepad++ it says it’s too big to open.
I’m sorry for the questions.

Sign up or log in to comment