Deep-learning-activity-recognition
Deep-learning-activity-recognition copied to clipboard
Bug in code
Hello, you have a bug in your code, when you use nn.CrossEntropyLoss() it automatically applies softmax function, so you should input just raw outputs,but instead you input vector in softmax layer and then you apply it to nn.CrossEntropyLoss() loss, that is mistake. here is link: https://github.com/jindongwang/Deep-learning-activity-recognition/blob/6ec30f24c54f47d30fbaae875a9cd40aba256442/pytorch/main_pytorch.py#L23
@datonefaridze Much thanks for pointing this out! It is true that I shouldn't use F.softmax
before nn.CrossEntropyLoss
! I'll update the code.