NVIDIA / OpenSeq2Seq

Toolkit for efficient experimentation with Speech Recognition, Text2Speech and NLP
https://nvidia.github.io/OpenSeq2Seq
Apache License 2.0
1.54k stars 369 forks source link

19.01 transformer #339

Closed borisgin closed 5 years ago

borisgin commented 5 years ago

Added two new experimental features: 1) batchNorm and layerNorm with L1 options (layernorm with L2 is default) 2) L2 regularization for encoder and decoder (default None) Both options are backward compatible with previous config files and models