Skip to content
This repository has been archived by the owner on Oct 31, 2023. It is now read-only.

Degraded DCv2 Training Loss #100

Open
gorkaydemir opened this issue Mar 29, 2022 · 0 comments
Open

Degraded DCv2 Training Loss #100

gorkaydemir opened this issue Mar 29, 2022 · 0 comments

Comments

@gorkaydemir
Copy link

Hi,
Congratulations on your great work and thank you for making it public.
I am training DCv2 on my custom dataset, however training (Cross Entropy) loss goes to zero in strictly decreasing manner. Moreover, it's linear evaluation performance degrades in each epoch. What could be the reason?
I have also tested with additional entropy term to prevent possible collapsing (like in swaw) but it did not help.

Thanks in advance.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant