Loss value

[CIntellifusion]

hey thanks for your implementation.
The code works well.
But do you have any idea about the loss value which is more than 100?
Is it due to the KL divergence?

[satyrmipt]

Loss value in VAE is a sum of reconstruction loss (BCE in this repo) and KLD (between learnable distribution and standard normal distribution in this repo). Non of them has to be less then 100.