Open sujeendran opened 4 years ago
The expected good loss value for a reasonable result is below -6.0. And according to my experience, after the training loss is below -6, the loss decrease extremely slow, but it does decrease. And the more you train, the better the audio is generated.
Hi! Can you help me understand how is the loss calculated and what is the expected good loss value from training the model?