search

grey-eye / talking-heads

Our implementation of "Few-Shot Adversarial Learning of Realistic Neural Talking Head Models" (Egor Zakharov et al.)
GNU General Public License v3.0
593 stars 110 forks source link

Training issues #38

Open gozuslayer opened 5 years ago

gozuslayer commented 5 years ago

Despite having only one GPU, I was able to test your code on batch size of 2 instead of 3. But I have an issue with the losses, each one (EG and D) becomes NaN ... Any clues?

danmarbeck commented 5 years ago

Have a look at #26. This should be the problem you are encountering.