Do not multiply loss by target's batch size

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 multiply loss by target's batch size #722

Closed mannatsingh closed 3 years ago

mannatsingh commented 3 years ago

Summary: This is unnecessary and we unscale these values in our hooks later on. More importantly, this adds the assumption that the target is a tensor which this diff gets rid of - https://github.com/facebookresearch/ClassyVision/issues/719

Differential Revision: D27212190

facebook-github-bot commented 3 years ago

This pull request was exported from Phabricator. Differential Revision: D27212190

facebook-github-bot commented 3 years ago

This pull request was exported from Phabricator. Differential Revision: D27212190

facebook-github-bot commented 3 years ago

This pull request was exported from Phabricator. Differential Revision: D27212190