No snapshots duirng training

NVlabs / stylegan3

Official PyTorch implementation of StyleGAN3

Other

6.38k stars 1.12k forks source link

No snapshots duirng training #197

Closed JulesPad closed 2 years ago

JulesPad commented 2 years ago

Describe the bug While training Stylegan2 on the ffhq data set with a 128x128 resolution on Pytorch, I have only the first 3 training snapshots. I am currently on tick 22, and have only the snapshot of tick 000001 and 000002.

Screenshots status output

Desktop (please complete the following information):

Linux Ubuntu 20.04
PyTorch 1.10.1
CUDA 10.1
Rtx 2060 6Gb
Docker: did you use Docker? No

JulesPad commented 2 years ago

Just found out that snapshots are produced every 50 ticks. It is possible to find this and modify this value in the train.py file.

I have seen some other people having the same problem, hope this will help.