search

MicrosoftDocs / ml-basics

Exercise notebooks for Machine Learning modules on Microsoft Learn
MIT License
1.76k stars 2.27k forks source link

The torch deep learning model should not use softmax last layer + CrossEntropyLoss #48

Closed yxu02 closed 3 years ago

yxu02 commented 3 years ago

Torch CrossEntropyLoss expects logits instead of normalized probabilities.

GraemeMalcolm commented 3 years ago

Thanks - fixed