Closed Bachstelze closed 4 years ago
Oh, interesting how that slipped through. We had CPU translation (and training and scoring) in Marian since early 2018. Thanks for pointing this out. Just use --cpu-threads N
to parallelize over input batches.
OK, thanks for the quick response.
Fixed. Thanks!
The current documentation declares different states for the cpu translation.
Or do I confuse it with amun?