search

SKTBrain / DiscoGAN

Official implementation of "Learning to Discover Cross-Domain Relations with Generative Adversarial Networks"
773 stars 170 forks source link

Face Training Time Too Long #4

Open dakami opened 7 years ago

dakami commented 7 years ago

Training faces is set up for 5,000 epochs, which on my Pascal Titan X requires a few minutes per epoch. There are three things I can imagine have gone wrong:

1 (and most likely): The dataset I'm using comes from the Google Drive link, and has more files than you were training against 2: The images in the dataset need to be downsampled 3: Epoch default is too high

Or maybe something else?

jazzsaxmafia commented 7 years ago

3 seems to be the reason. You don't have to train the model up to all 5000 epochs to get decent results.

dakami commented 7 years ago

Maybe update the docs? What's a good amount for this dataset?

d2sys commented 6 years ago

I' training the model on celebA data and it takes about 12 minutes for a single epoch I have GeForce GTX 1050 4GB