YuyingShang
Results
2
issues of
YuyingShang
when I run the code, no matter lsr-bert model or lsr model, after the evaluate_epoch=30, my process will be killed. For both models, I have set my batch_size=3, lr=1e-3. Do...
我注意到在Repadapter的论文中,实验表明在FFN和MHA前都加入适配器可以获得更好的效果。但是在Lavin的训练代码中,我注意到您仅在MHA前应用了适配器,请问为什么舍弃在FFN前添加适配器的操作了呢?