execute python generate_fakes.py --checkpoint files/vatr.pth have wrong

[yuyukuy534]

Hello, I get this error "assert len(old_model) == len(checkpoint) " when I execute python generate_fakes.py --checkpoint files/vatr.pth code, what is the reason for this?

[vittoriopippi]

I've just cloned the repo, and the command python generate_fakes.py --checkpoint files/vatr.pth doesn't give me any error...

[yuyukuy534]

Hello, @vittoriopippi can you send me your generate_fakes. py file, I downloaded it from github and run it all the time with AssertionError

[vittoriopippi]

The downloaded file I've used is what you can find on GitHub. Which checkpoint are you using for the generation?

[yuyukuy534]

Hello,@vittoriopippi I downloaded vatr.pth and command python generate_fakes. py--checkpoint files/vatr.pth, reporting an error of AssertionError

[vittoriopippi]

There is probably something wrong with your checkpoint since the vatr.pth checkpoint I've shared works completely fine. You are probably using a checkpoint trained by yourself. In that case, you should change the arguments of generate_fakes.py to match the same ones you've used during training.

If you have downloaded it, please ensure the checkpoint is downloaded correctly. Check if the checksum is the same as below:

(vatr) user@node:~/VATr$ md5sum files/vatr.pth 
d1386fd46314dd281d9046cd4d6e3852  files/vatr.pth

If not please download it again (link)

[yuyukuy534]

@vittoriopippi Thank you very much for your timely reply. After I downloaded vatr.pth correctly, I executed the command to report an error of 'KeyError:' img_id '. What caused this?

[vittoriopippi]

Please follow the installation guide in the readme because you are using the wrong files. The IAM-32.pickle I've used differs from the one shared by HWT. My pickle file also includes a unique ID for each image.

https://github.com/aimagelab/VATr#installation

[yuyukuy534]

Ok, thank you very much for your reply and guidance

[yuyukuy534]

@vittoriopippi Hello, I have been looking up information to learn how to use OOV_word.text to calculate the FID value of OOV, but I still don't understand it. I hope I can get your guidance and help