Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Update figure_iou.py #387

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Update figure_iou.py #387

wants to merge 1 commit into from

Conversation

salkhokhar
Copy link

@salkhokhar salkhokhar commented Aug 1, 2022

As mentioned in issue (#319), sometimes during training the IOU loss gets stuck at nan. @qinhao14 suggested to use the function torch.nan_to_num() and it did solve the problem for me. And after multiple training sessions, no new problem seems to have arisen because of this.

As mentioned in issue [meituan#319] (meituan#319), sometimes during training the IOU loss gets stuck at nan. [@qinhao14 ] (https://github.com/qinhao14) suggested to use the function torch.nan_to_num() and it did solve the problem for me. And after multiple training sessions, no new problem seems to have arisen because of this.
@shensheng272
Copy link
Collaborator

Thanks for the pr.
Torch.nan_to_num() did make training more robustly on large batchsize and custom data. However most of our exps did not use this trick. All release model also did not use this trick.
So could you change code so torch.nan_to_num() check is off by default. Like store true training params

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

Successfully merging this pull request may close these issues.

2 participants