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.

Sign up for GitHub

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

Jump to bottom

loss issue in RegL1Loss #693

Closed

JavisPeng wants to merge 1 commit into xingyizhou:master from JavisPeng:loss_issue

JavisPeng commented Apr 25, 2020

def forward(self, output, mask, ind, target):
    pred = _transpose_and_gather_feat(output, ind)
    mask = mask.unsqueeze(2).expand_as(pred).float()
    # loss = F.l1_loss(pred * mask, target * mask, reduction='elementwise_mean')
    loss = F.l1_loss(pred * mask, target * mask, size_average=False)
    loss = loss / (mask.sum() + 1e-4)
    return loss

since the mask has expanded, the number of value 1 should be twice the previous, which results in loss=loss*0.5


          loss issue

19327ca

Owner

xingyizhou commented May 13, 2020

Thank you for your contribution. However this will impact the performance and needs further verification. Closing it for now.

xingyizhou closed this

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet