Error more than 10000 in reconstruction error!

[bhavithran1]

Should the reconstruction error be this big and not low

Checked on test dataset and my dataset and the results are just really big

[bhavithran1]

This might be caused because of nn.MSELoss(reduction=‘Sum’)

Should I change it to mean because that sounds more better with kL being more lower when a sigmoid layer is applied to the encoder and decoder