After thoroughly readinging your codes, I'm wondering how you perform model selection.
Did you perform model average on the several last checkpoints? Or did you evaluate each checkpoint on the validation set and them average the several best checkpoints?
I did not find any clue in the codes. Any reply will be appreciated, thanks!
Thanks for the great work.
After thoroughly readinging your codes, I'm wondering how you perform model selection. Did you perform model average on the several last checkpoints? Or did you evaluate each checkpoint on the validation set and them average the several best checkpoints?
I did not find any clue in the codes. Any reply will be appreciated, thanks!