NAN values in losses while training

weiyithu / SurroundDepth

[CoRL 2022] SurroundDepth: Entangling Surrounding Views for Self-Supervised Multi-Camera Depth Estimation

MIT License

247 stars 38 forks source link

NAN values in losses while training #22

Open aaron-hgx opened 1 week ago

aaron-hgx commented 1 week ago

While I am training with scale aware pretraining, I am getting NAN values in losses. What could be the possible issue? Without scale aware pretraining the losses seem to be fine.