ValueError: Image input is required in the image2video pipeline. Please provide a regular image file (image_file = None).

[dieterporsche]

I have adjusted the yaml file and i am still using the provided data. But it wont work it seems like you have a bug:

ValueError: Image input is required in the image2video pipeline. Please provide a regular image file (image_file = None).

[sixsixcoder]

You need to provide a real image path

[TousakaNagio]

Hi @dieterporsche,
Have you solved the issue? I ran into the same issue while fintuning the i2v model.

[dieterporsche]

Hi @TousakaNagio
Unfortunately not yet. The finetuning worked fine a couple weeks ago, i am not sure yet what the problem is now. I also provided the real image path but i still run into the problem.

[TousakaNagio]

Have you reinstalled the environment? I tried to install it recently and ran into the issue.
In my case, I have to comment lines 73-75 in site-packages/cogkit/utils/misc.py to bypass the error.
But I am not sure if it is a feasible solution.

[Zane-ZYQiu]

Have you fixed it? I got the same problem.

[dieterporsche]

@Zane-QIU Hello. Yes, I’ve fixed it based on @TousakaNagio comment. You basically need to comment out lines 73–75 in the site-packages/cogkit/utils/misc.py file. However, every time you run:

pip install "cogkit@git+https://github.com/THUDM/CogKitFix.git"
you’ll need to comment out lines 73–75 again.

I’ve also created a fork where this is already fixed, so you can instead run:

pip install "cogkit@git+https://github.com/dieterporsche/CogKitFix.git"
That should resolve the error.