This is the code for ACL Findings Paper Encoder and Decoder, Not One Less for Pre-trained Language Model Sponsored NMT.
This code is based on Fairseq 0.10.2.
The version of transformers is 4.29.2.
Usage:
- python fwet-embed.py
- java bpe
- python fairseq_cli/preprocess.py
- python fairseq_cli/train.py
- python fairseq_cli/generate.py
Results of Chinese-English will be uploaded to github and arxiv.
Results of model using different PLMs on WMT14 English-German will be uploaded to github and arxiv.