More details on models?

[shuoyangd]

Hi,

I'm trying to reproduce the models that are provided as part of the shared task. However, I found the hyper-parameter settings for these models were only vaguely described in the paper (was looking at this one).

Is it possible to provide more details on that? Most desirably, the exact fairseq command used to train these models will be very helpful.

Thanks!

Shuoyang

[shuoyangd]

Ah, just realized that I can recover these arguments from the model dump. The model dump is a dict with an entry args.