search

Justin-Tan / high-fidelity-generative-compression

Pytorch implementation of High-Fidelity Generative Image Compression + Routines for neural image compression
Apache License 2.0
411 stars 77 forks source link

Total steps of training #16

Closed amruz closed 3 years ago

amruz commented 3 years ago

Thank you for the repository! Regrading the total number of epochs of training, from the original paper they are training both part (autoencoder/hyperprior and GAN traininig) for 1 million steps. However in this implementation its trained for 10 epochs (If I am not wrong). Could you please explain this

Justin-Tan commented 3 years ago

Hi, https://github.com/Justin-Tan/high-fidelity-generative-compression/blob/master/train.py#L232

Optimization stops at the earlier of 1e6 steps/10 epochs. If you measure progress in epochs, this will depend on the size of your training set.

amruz commented 3 years ago

Thank you!