sxela / discodiffusion-warp Goto Github PK

fix whitespaces in filenames error
some paths and filenames won't work with whitespace. find and shield all shell commands' paths

Clear messy defaults like leaving resume run in production.

I love the idea behind this built, thanks for sharing.

I'm having trouble with "Define optical flow functions for Video input animation mode only"

First error was
[Errno 2] No such file or directory: '/content/RAFT/core'

So I changed it to make sure it found the folder with root_dir ='/content/drive/MyDrive/AI/Disco_Diffusion'

Then I get a new error:

/content/drive/MyDrive/AI/Disco_Diffusion/RAFT/core
---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
[<ipython-input-45-c71df2f6329e>](https://localhost:8080/#) in <module>
     10   get_ipython().run_line_magic('cd', '{path}')
     11 
---> 12   from utils.utils import InputPadder
     13   from raft import RAFT
     14   # from utils import flow_viz

ModuleNotFoundError: No module named 'utils'

This No module named 'utils' is dumbfounding me, I see a utils folder but don’t how to fix this. I tried reinstalling, tried subbing in code from "Disco Diffusion v5.61" but had no luck. Any thoughts, I would really acpprieactee it. Thanks!

Add

torch.cuda.empty_cache()
import traceback
try:
  do_run()
except KeyboardInterrupt:
    pass
except Exception as e:
    print(e)
    print(traceback.format_exc())
finally:
    print('Seed used:', seed)
    gc.collect()
    torch.cuda.empty_cache()```

Do you have a python file format instead of ipynb which can be executed from command line? If you have please share it through GitHub. Thanks in advance!

create video cell:

frames_in = (glob(batchFolder+f"/{folder}({run})_*.png"))

should be

frames_in = sorted(glob(batchFolder+f"/{folder}({run})_*.png"))

[Errno 2] No such file or directory: '//content/drive/MyDrive/AI/Disco_Diffusion/RAFT/models/raft-things.pth'

pretty new to this. am i missing a file? maybe i deleted, if did how do i redownload it?

No matter what I do, overlapping is an issue. The only fix is in this colab which was an upgrade to yours but the person didn’t fix the local install issue. So can you add his overlap and maximize feature? I tried coding this in but I’m a novice.
https://colab.research.google.com/github/WASasquatch/discostream/blob/main/DiscoStream.ipynb

I tried your progrockdiffusion and the main issue of it not producing the prompt I’m putting in, it won’t let me change the ETA or clamp max. So can you fix your colab with gobig maximize and gobig overlap that’s in once again this colab https://colab.research.google.com/github/WASasquatch/discostream/blob/main/DiscoStream.ipynb

then add it to yours ASAP please

Add flo folder check before generating flows

add error message for bad video path when running video settings cell

I am trying to run Disco Diffusion on windows locally. I am getting following error.

0it [00:00, ?it/s]
ERROR: 0 flow files found.
Please rerun the flow generation cell.

replace with autodetected content

Hi,
it's more a question than an issue..
I'm wondering if motionvector images (creates by 3d-applications as Blender, Maya, etc) could be used as as Flow maps.
The basic idea is: I create a simple animation in a 3D-app which will be used as video-input, together with the vector pass as flow input.
I'm not sure if this would improve the results, but the flowmaps would be 100% accurate and pixelperfect.
So the the basic question would be, if you think the vector-images could be translated/converted to your flow-map format?
thx!