Oleksandr Kushnir comments

Repositories
Issues
Comments

Results 1 comments of


                                            Oleksandr Kushnir

trafficstars

Loss becomes nan after certain epoch.

It looks like I found one of the reasons for this issue. Let's look at this piece of code in [mlsd_pytorch/data/wireframe_dset.py](https://github.com/lhwcv/mlsd_pytorch/blob/main/mlsd_pytorch/data/wireframe_dset.py): ![image](https://user-images.githubusercontent.com/17236032/177992855-ee42607a-28fb-4d87-b367-5d3b9bdb2dfa.png) Providing that the input size is 512x512, junction_map...