Key hyperparameters to reproduce results？

[VickiCui]

Thanks for your great work! I'm having some trouble when reproducing the results. Although trained as in the tutorial, there is still a big gap between the model I trained and the results given in the paper. Can you provide the necessary hyperparameters in order to reproduce the results? Such as batch size, epoch, learning rate, beam size, and so on. Thx!!!

[tuetschek]

@VickiCui I don't think this is the correct place – the repo here is just for the metrics scripts, it doesn't include the baseline submissions. Perhaps @sebastianGehrmann or @yjernite would know more on the baselines?

[VickiCui]

您好，您的邮件我已经收到，我会第一时间处理。

[sebastianGehrmann]

Which task are we talking about? Most of the models were trained for https://openreview.net/forum?id=CSi1eu_2q96 and the hyperparameters from that paper should replicate results.