AliaksandrSiarohin

Without adversarial it is more stable.

40 is too large. Batch size should be approximately 4 times smaller than for 256, e.g. 16 or 12.

Probably just slow. You can try to change num_repeats to 1 to see. Also you may want to start with pretrained 256 checkpoint to accelerate convergence.

Depends on how much training you can afford. The more the better.

Since you tried it you should know better how much time it takes.

No idea, you can try. In principle there is nothing cuda specific.

This code does not belong to this repository, why you posting it here? You specify wrong path most likely or don't have images in your folder.

I have no idea, since you did not provide your input.

The model is trained on interview video sequences, I doubd such a movement exist in a interview videos. So you have to train on video sequence that is similar to...

Increase the number of ecpochs in the config and add ```--checkpoint vox-256.pth.tar``` to the train command.