NaN while training GAT model

apuaaChen / gcnLib

10 stars 4 forks source link

Open fishmingyu opened 2 years ago

fishmingyu commented 2 years ago

We run your code under the following setting:

python training_main.py --data Reddit --mode gas --model GAT
python training_main.py --data Reddit --mode gar --model GAT

However, we get nan in loss and test ACC 0.058, which turns out to be the wrong answer. Have you tested in such a mode? Hope to get your reply soon.

fishmingyu commented 2 years ago

You could check our screenshot here.