Neural Style Transfer - error with pretrained weights while using it on colab

I was trying to use the neural style transfer in my local machine for which I tried using the same code with same versions of dependencies as provided in the course lab in jupyter notebook. In the jupyter notebook everything went smoothly but the problem was that I have integrated graphics and therefore it took almost 1.5 hours to train the model 50% so I thought I could use google’s colab which provides gpu. But now I am getting an error while loading the weights which is:
OSError: Unable to open file (truncated file: eof = 11534336, sblock->base_addr = 0, stored_eof = 80134624)

I tried to load the weights from the pretrained-model folder as well as from the root folder as provided in colab
This is the whole error:

I re-uploaded the weights few times but still no luck. I refreshed and re-ran the the run time but again it didn’t work. Please help.

Note that all the notebooks in the courses here use versions of TF and other packages that are pretty out-of-date w.r.t. the currently published versions of everything. Things change really quickly in this space, so you can’t assume that code that worked with the April 2021 version of TF and the other packages will “just work” with the current versions.

Maybe you should run the training also on Google Colab, so that you get the weights saved in a compatible format. Or if you mean the VGG weights, you’ll need to look around for a version that is compatible with whatever version of TF you are running.

I installed the same version of TF in the colab which was used in the course lab and also everything including weights were working fine when used in my local jupyter notebook.
I think I can find the VGG weights for latest TF but still I would be curious what could be causing the error.

It’s an interesting question, but it has nothing to do with the course materials here. Please realize that the mentors are volunteers here: we do not get paid to do this. Which means that we have discretion about how we choose to spend our time here. There is nothing to prevent someone from further investigating your question and helping you, but my personal choice is that I would rather invest my time helping students with questions about the course materials. If you learn anything interesting about this in your future investigations, you are welcome to share the knowledge by posting it here and maybe it will help future visitors here who have a similar question.


I totally understand it, Paul. Thanks for helping out. I will certainly share the solution as soon as I find it, so others can make use of it.