Hi,
You mention in the paper that you have excluded the next-sentence prediction objective from XLNet since it didn't introduce any improvements, However in the Ablation study you also report the performance in case of using NSP.
My question is : is NSP implemnted here in your github repo or not?
Hi, You mention in the paper that you have excluded the next-sentence prediction objective from XLNet since it didn't introduce any improvements, However in the Ablation study you also report the performance in case of using NSP. My question is : is NSP implemnted here in your github repo or not?
Thanks a lot