DUL

DUL copied to clipboard

KL-loss is too large

3

Thanks for your great work! I noticed the KL-loss in your log is about 4. But when I run this code, the KL-loss was too large and turned out to...

zerooooone

关于标准差的范围

1

你好。原文中提到将σ限制到[0,1]范围内，以保证KL损失在单调递减范围内，但是我在代码里面好像没有看到有限制std的范围，请问是什么原因？

asd5152306

Checkpoints

2

Hi, Are checkpoints available for this model or should I train the model to get them? Thanks.

HOMGH

关于代码的问题

1

您好我想问一下，这个模型在测试阶段还需要使用方差吗？在测试阶段应该直接用mu就行了吧，但是我看代码好像是都使用了

pswena

It seems like you did not use softmax loss to classification as the paper said. [This class](https://github.com/MouxiaoHuang/DUL/blob/860a08178037eca4aa919dbb1ae3f101fb1716ab/head/metrics.py#L60) You just used a fully connected layer, so it is another CEloss. Is...

PeiwenSun2000