Fix a bug when resuming training

ZhengPeng7 / BiRefNet

[CAAI AIR'24] Bilateral Reference for High-Resolution Dichotomous Image Segmentation

https://www.birefnet.top

MIT License

322 stars 28 forks source link

Fix a bug when resuming training #27

Closed Alive1024 closed 2 months ago

Alive1024 commented 2 months ago

Hi, thanks for your great work! I found a minor bug when resuming training. In train.py’s init_models_optimizers function, epoch_st needs a global declaration to make resuming work properly.

ZhengPeng7 commented 2 months ago

Thanks, you are right. I seldom use the resuming and neglected this bug. I'm merging this PR.