VSGAN - RealESRGAN Compact model crash

[TheLastApostle]

Hello,

I am currently trying to use Sudo's Ultracompact model: <https://www.mediafire.com/folder/7e7752gf42eky/sudo_UltraCompact_2x_1.121.175> in order to upscale some anime footage.

Sadly, the moment I try to upscale anything I am prompted with an immediate crash. Is there a fix for this?


Another question.
My input is 1920x1080, is there a way to resize it before the upscaling pass to 1280x720 in order to speed the process up? Crop/Resize only works on the output.

Thank you

[Selur]

Another question.
My input is 1920x1080, is there a way to resize it before the upscaling pass to 1280x720 in order to speed the process up? Crop/Resize only works on the output.

Setting "Filtering->Vapoursynth->Misc->Script->Lower res. before resize" should allow this.

Sadly, the moment I try to upscale anything I am prompted with an immediate crash. Is there a fix for this?

Not unless VSGAN starts to support these models, see: https://github.com/rlaphoenix/VSGAN/issues/25
When you get something like:

Failed to evaluate the script:
Python exception: max expected at least 1 argument, got 0

Traceback (most recent call last):
File "src\cython\vapoursynth.pyx", line 2866, in vapoursynth._vpy_evaluate
File "src\cython\vapoursynth.pyx", line 2867, in vapoursynth._vpy_evaluate
File "G:\Temp\tempPreviewVapoursynthFile04_29_52_850.vpy", line 65, in
vsgan.load(model)
File "I:\Hybrid\64bit\Vapoursynth\Lib\site-packages\vsgan\networks\esrgan.py", line 46, in load
model = arch(state)
File "I:\Hybrid\64bit\Vapoursynth\Lib\site-packages\vsgan\archs\ESRGAN.py", line 72, in __init__
self.num_blocks = self.get_num_blocks()
File "I:\Hybrid\64bit\Vapoursynth\Lib\site-packages\vsgan\archs\ESRGAN.py", line 228, in get_num_blocks
return max(*nbs) + 1
TypeError: max expected at least 1 argument, got 0

this usually means that VSGAN doesn't support the model format. (https://vsgan.phoeniix.dev/en/stable/#supported-models)

As a side note, when using different models, also make sure to rename them for example from sudo_UltraCompact_2x_1.121.175_D.pth to 2x_sudo_UltraCompact_1.121.175_D.pth since this will allow Hybrid to know the scaling factor (2x). (iirc Hybrid uses 2x by default if the name doesn't start with Yx, so it's not important in this case, but it might be in others)

Cu Selur

[TheLastApostle]

Hi,

That's really depressive, I was finally getting my hopes up that I may be able to have an alternative to the nightmare that VSGAN Docker can often be to handle.

I assume this is basically a dead end, VSGAN has not had any 'upgrades' in nearly an year now so I reckon the future of it isn't bright either.

Thank you anyway,
Have a nice evening.

[Selur]

btw. problem seem to be 'Compact'-models while 'RealESRGAN Compact'-models like https://github.com/Bubblemint864/AI-Mode...Compact_v1 seem to work fine.

Cu Selur