Loss is nan, When I do not provide actually label

XifengGuo / DEC-DA

Deep Embedded Clustering with Data Augmentation (DEC-DA). Performance on MNIST (acc=0.985, nmi=0.960).

MIT License

55 stars 20 forks source link

Open TeerathChandani opened 5 years ago

TeerathChandani commented 5 years ago

Hello

I tried to train on my own dataset. It is working fine, when I provide the actual (y). When y is None. its loss is nan.

So please fix it or tell me, where I am doing wrong?

Thanks