5 ms·

I guess by posting on Facebook? It's not clear from the post whether they are running this yet. Otherwise maybe https://huggingface.co/ https://huggingface.co/

• Jack000 6y ago
  note that the fairseq implementation requires 2x 32GB v100 gpus for inference.
  • visarga 6y ago
    And that's why we won't see this deployed in production anywhere.