Question regarding accessing output

[Jana Vozárová]

Hello,

I would like to perform a manual analysis of the translated sentences (in case of eng-cs data).  Is there a command/script that lets you view or further access the output/translated text? Such as producing a .txt file in which the translations are stored?

Thank you and best regards,

Jana

[Rico Sennrich]

Hello Jana,

you can use the translate.py script to generate translations of any input file. It writes to stdout by default, but you can provide the "-o" argument to define an output file.

if you're asking about the translations generated during training/validation, they're written to a temporary file that is afterwards deleted. You could change this behaviour in train.py (function validate_with_script() ), or simply store checkpoints and then use the translate.py script to re-translate your validation/test sets.

best wishes,

Rico

--
You received this message because you are subscribed to the Google Groups "Nematus Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nematus-suppo...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/nematus-support/CAOWgqAMswHd59Nws-VzxNEfESbS057tJ9qijYX%2Bze8pT5ALO%3Dw%40mail.gmail.com.