You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am trying to replicate the DiffuSeq model for the Paraphrase task with the QQP dataset. I kept everything to the default training config, and for MBR I ran with 20 different random seeds during evaluation, but I still can't match the performance reported in the paper in Table 1.
For reference, this is the content of training_args.json file:
Hi!
I am trying to replicate the DiffuSeq model for the Paraphrase task with the QQP dataset. I kept everything to the default training config, and for MBR I ran with 20 different random seeds during evaluation, but I still can't match the performance reported in the paper in Table 1.
For reference, this is the content of training_args.json file:
and this is the output of running evaluation
python eval_seq2seq.py --folder ../{your-path-to-outputs} --mbr
:The text was updated successfully, but these errors were encountered: