VAE Training Details

[ahmed1996said]

Dear Authors,

Thanks for the great work, and for making the repository public!

I'm training my own VAE on PadChest dataset, and I was wondering if you can provide the hyperparameters used in training? Can you confirm whether or not the hyperparameters listed on this notebook are correct? In the paper, it was mentioned that both frontal and lateral images were used in VAE training, but in the notebook I can see these args:

'frontal_only': 1,
'val_frontal_only': 0,
'ignore_nonfrontal_loss': 1,

Also, can you provide any details about the final performance of the VAE (SSIM, Reconstruction Loss, etc.) so I can ensure I'm reproducing the model correctly?

Thanks in advance!