@neuralmagic
Neural Magic
2 years
BERT-Large (345M parameters) is now faster than the much smaller DistilBERT (66M parameters) with accuracy of the larger BERT-Large! It delivers 8x latency speedup on commodity CPUs 🚀 🙏 @ZafrirOfir , @guybd35 , @markurtz_ & teams for fantastic collab!
Tweet media one
0
8
37