search

dongxingning / SHA-GCL-for-SGG

Code for paper "Stacked Hybrid-Attention and Group Collaborative Learning for Unbiased Scene Graph Generation"
MIT License
32 stars 5 forks source link

Optimizer in code is SGD, not Adam #9

Open Git-oNmE opened 2 years ago

Git-oNmE commented 2 years ago

Hello, appreciate for your solid work first.

We saw in paper that this model is optimized by Adam, while we check the code and only find SGD optimizer. image

We tried to change the optimizer by ourselves, but failed. image image

While if we use the SGD optimizer as origin code gives, the code can work. image image

So how can we use Adam in this code? Looking forward to your reply, thanks!