ProGan RuntimeError about machine-learning-collection HOT 9 CLOSED

Yeah sorry, I didn't push the latest version that I was working with. It should work now

from machine-learning-collection.

thanks !

from machine-learning-collection.

@aladdinpersson now i got RuntimeError: CUDA out of memory. i have Intel core i7 NvidiaGeoforce 940 mx CUDA 11.1

from machine-learning-collection.

Change the image size in config file from 128 to something lower like 4/8/16

from machine-learning-collection.

@aladdinpersson that's strange even with image size 4 it does output same error. Maybe a bug in (Pytorch pip windows 10 version) ?

from machine-learning-collection.

From my understanding your gpu only has 2gb of vram which is probably the reason. You can try to reduce the number of channels in the config file and see if that works. Since we're creating all the conv layers etc for up to 1024x1024 generation you do need to have some vram because the parameters will be on the GPU. I haven't done any extensive tests but it runs on my GTX 1060 with 6gb vram

from machine-learning-collection.

4gb of vram ,but it is okey, i will go on colab . Thanks!

from machine-learning-collection.

@aladdinpersson sorry for bothering but in my colab i get this error:

RuntimeError: Given transposed=1, weight of size [100, 256, 4, 4], expected input[8, 256, 1, 1] to have 100 channels, but got 256 channels instead

from machine-learning-collection.

well i fixed it with Z_DIM = 256

from machine-learning-collection.