Porting fairseq wmt19 translation system to transformers

The article discusses the successful porting of the fairseq WMT19 translation system to the Hugging Face Transformers library. Key technical details include the adaptation of the original sequence-to-sequence architecture, enabling compatibility with the Transformers framework while maintaining performance benchmarks. This transition allows practitioners to leverage the extensive ecosystem of Transformers, facilitating easier integration and experimentation with state-of-the-art translation models.