You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Jul 7, 2023. It is now read-only.
Hello
I am using tensor2tensor (1.2.9),and I am not sure if training is in a correct way . When I do decoding , error occurs (it always truncate the source sentences by symbol "."):
Examples are listed as follows :
My configuration is transformer_base_single_gpu , transformer and all others are set to default .
The bleu score only get 12.16 in newstest2014 (mirror question #317 ).
The text was updated successfully, but these errors were encountered:
Why do you create a new issue for the same problem as in #317?
As I wrote there, I believe the BPE vocab problem was fixed in t2t 1.3.0 by 7909c69.
You wrote there you will try it with t2t 1.3.0, but now you are reporting 1.2.9 again (or is it a typo?).
@martinpopel Thanks,I am running the program t2t 1.3.0 . Before getting the result , I printed the translation log to find the error occurs . I am afraid u can not receive the message in time , so a new issue is pulled . Sry ~
Hello

I am using tensor2tensor (1.2.9),and I am not sure if training is in a correct way . When I do decoding , error occurs (it always truncate the source sentences by symbol "."):
Examples are listed as follows :
My configuration is transformer_base_single_gpu , transformer and all others are set to default .
The bleu score only get 12.16 in newstest2014 (mirror question #317 ).
The text was updated successfully, but these errors were encountered: