search

codeslake / IFAN

[CVPR 2021] Official PyTorch Implementation for "Iterative Filter Adaptive Network for Single Image Defocus Deblurring"
GNU Affero General Public License v3.0
227 stars 39 forks source link

Loss NAN #16

Closed TPZZZ closed 1 year ago

TPZZZ commented 1 year ago

hello Junyong Lee: Thanks to your great work! I got a problem of Loss nan when I retrained the IFAN using my training setting, without using the losses reblurring and defocus map estimation by duai pixel data. Do these losses make the loss be nan during the training? Thank you for your help! @codeslake

codeslake commented 1 year ago

You can try to decrease your learning rate or gradient clipping value.