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
A document with 50k words takes about a day to translate on a 16 year old machine. For the past day, I’ve been working on a doc double that size. There is no progress indicator and the target output file is zero in size, so I have no idea how far along the processing is. If I lose power or need to shutdown the machine, the past day of work is lost. Most people have newer hardware but considering the heavy amount of computation needed anyone can be trapped in a task that takes days given enough text to translate.
Workaround: we can split the original doc into manageable pieces. But this is manual labor intensive because we cannot just chop the doc at arbitrary locations. Every piece must not break a paragraph. Then the pieces can be in such high numbers that it’s hard to manage.
So I suggest these improvements:
A progress indicator
A way to send a signal to stop processing gracefully and preserve the work completed so far
A way to resume processing on a giant document later
A way to see and make use of partial translations. E.g. on day 5 of a translation process, a user should be able to access output generated 2 days into the translation.
The text was updated successfully, but these errors were encountered:
A document with 50k words takes about a day to translate on a 16 year old machine. For the past day, I’ve been working on a doc double that size. There is no progress indicator and the target output file is zero in size, so I have no idea how far along the processing is. If I lose power or need to shutdown the machine, the past day of work is lost. Most people have newer hardware but considering the heavy amount of computation needed anyone can be trapped in a task that takes days given enough text to translate.
Workaround: we can split the original doc into manageable pieces. But this is manual labor intensive because we cannot just chop the doc at arbitrary locations. Every piece must not break a paragraph. Then the pieces can be in such high numbers that it’s hard to manage.
So I suggest these improvements:
The text was updated successfully, but these errors were encountered: