Starting loss is too high.

say4n / pytorch-segnet

SegNet implementation in Pytorch framework

MIT License

88 stars 35 forks source link

Starting loss is too high. #7

Closed farhadinima75 closed 3 years ago

farhadinima75 commented 3 years ago

hi say4n.

tnx for your nice and simple implementation.

I've encountered a simple issue in the first epoch loss.

starting loss is high (nearly 150) for me.

is it ok? does this happen for you when testing your algorithm?

any answer would be appreciated.

say4n commented 3 years ago

Hi @farhadinima75 thank you for the kind words, however, I no longer have access to the lab where I worked on this. So, I cannot look up your query. 👍