Open Journey7331 opened 3 days ago
Hi, @tian-qing001 Thanks for your great work!
Is there any ablation study on the activation function? And why q + 1.0 after applying elu? ensure q > 0 for better training? 👀
q + 1.0
elu
q > 0
https://github.com/LeapLabTHU/MLLA/blob/5a1719682cb040ec708fb633ad1379c9eab28576/models/mlla.py#L129-L130
Hi, @tian-qing001 Thanks for your great work!
Is there any ablation study on the activation function? And why
q + 1.0
after applyingelu
? ensureq > 0
for better training? 👀https://github.com/LeapLabTHU/MLLA/blob/5a1719682cb040ec708fb633ad1379c9eab28576/models/mlla.py#L129-L130