Do not check for inf / nan loss in eval phase

facebookresearch / ClassyVision

An end-to-end PyTorch framework for image and video classification

https://classyvision.ai

MIT License

1.59k stars 278 forks source link

Do not check for inf / nan loss in eval phase #743

Closed mannatsingh closed 3 years ago

mannatsingh commented 3 years ago

Summary: The loss being NaN / inf during training is problematic since we cannot backprop using an invalid loss. This isn't the case during an eval phase and we don't need to crash.

Differential Revision: D28204666

facebook-github-bot commented 3 years ago

This pull request was exported from Phabricator. Differential Revision: D28204666