Speed-optimized, Compact Student Models that Distill Knowledge from a Larger Teacher Model: the UEDIN-CUNI Submission to the WMT 2020 News Translation Task.
Ulrich GermannRoman GrundkiewiczMartin PopelRadina DobrevaNikolay BogoychevKenneth HeafieldPublished in: WMT@EMNLP (2020)