Dmitry Ulyanov

Hello @piteight, this code is cpu only as far as I remember, so it cannot be caused by gpu. Looks like something unrelated to this particular repo. Try reinstalling TF...

Hi, thank you for reporting. A PR will be very helpful!

Hi Ianni, I am not an expert in TF, but in torch code there is a possibility to load a pretrained model. I am sure it is easy to change...

Hi, torch implementation of Instance Norm is easily hackable and can use stored mean/var instead of independent normalization at test time. I think you just need to set `self.bn` to...

Hello, I tested everything using 12GB card, so all the parameters tuned to work in my settings. You can try to decrease `batch_size` to 1 to see if it not...

did you generate hdf5 file first?

there's no sample hdf5 file, since it is too large. You should let it work till it finishes.

Try doing batch_size = 1, do not change ncolors, you can also downsize the image to 256x256 for example

The quality will be ok, I used batch_size = 1, but at test time you need to experiment with `midel:evaluate()` or `model:training()`

Did you try to update everything? Or redownload the file, it can be corrupted. I don't know other reasons why it could happen..