eric-haibin-lin opened a new issue #17507: Make loss scale public in AMP URL: https://github.com/apache/incubator-mxnet/issues/17507 For NLP often users need to do gradient norm clipping. When using AMP, I find it's not easy to access the current gradient scale factor unless I look into AMP source code: https://github.com/dmlc/gluon-nlp/blob/master/scripts/bert/finetune_squad.py#L464 Shall we make the loss scale a public property? @ptrendx
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
