Squeezing a little more memory

[aaronjolson]

The Problem I am having is that using some image combinations and settings my Python instance crashes and I get a message saying "Python has stopped working" I am trying to run python neural_style.py --content_img wolf.jpg --style_imgs electricity.jpg --max_size 1920 --device /gpu:0 --max_iterations 50 --verbose where wolf.jpg is 3500x3500px for 3.5mb, and electricity.jpg is 1920x1080px at 270kb

I am running on Windows 10 My gfx card is a Nvidia 1070 I have 16gbs of ram on my machine if that is relevant

The important thing to me is that my output images are sized 1920. Some image combinations work for me, others do not. I realize I am probably right on the edge of what my machine is capable of outputting.

What do I have to do to get some extra memory available? Are there any other settings I am not using that might give me some extra room?

[ferlix9o]

I would like to know if anyone manage to optimise the memory consumption as well.

would be cool to render 2K imgs.

+1

[Samreay]

Did either of you figure this out? Im still getting it falling over trying to request 5GB of VRAM for a max size 512 image using the adam optimiser, which was supposed to significantly reduce memory consumption.