why pred.exp()?

[SauceCat]

Thanks for the nice implementation 🐱 💯, I have a small question:

DDRNet.pytorch/lib/datasets/base_dataset.py

Line 222 in bc0e193

return pred.exp()

Why pred.exp() in inference? The real scores should be softmax(pred, dim=1) right? Just wonder what is the usage of adding exp() here, is it for visualization?