`predict.py` causing errors about gnnetworkingchallenge HOT 9 CLOSED

Hello!

I'll need some extra information to help you out:

• May you share the bash command where you call predict.py?
• May you share the entirety of the terminal output when calling predict.py?
• Have you done changes to the training or testing pipeline? If so, which?

from gnnetworkingchallenge.

Hi I am also facing the same problem

from gnnetworkingchallenge.

Hello,

I've tried to re-run the commands as described in the README and wasn't able to find the issue. Could you provide me with more details (i.e. the error message, if you made modifications to the code, which libraries did you use to build your solutions...).

Otherwise I will not be able to help you out.

from gnnetworkingchallenge.

Ok I have followed the commands in the readme.

The error is this : ValueError: Unable to load weights saved in HDF5 format into a subclassed Model which has not created its variables yet. Call the Model first, then load the weights.

The error is in the line: model.load_weights(args.ckpt_path) in predict.py

Even though the model is in the ckpt_path .. the files i have are :

from gnnetworkingchallenge.

The command I ran is : python3 /home/ec2-user/gnnet-ch23-dataset-cbr-mb/GNNetworkingChallenge-2023_RealNetworkDT/predict.py -ds CBR+MB --ckpt-path ckpt/Baseline_cbr_mb/best_08_10 --tr-path "gnnet-ch23-dataset-cbr-mb/extracted_08_10_cv/0/training" --te-path "/home/ec2-user/gnnet-ch23-dataset-cbr-mb/gnnet-ch23-toy-dataset/extracted_toy" --toy

from gnnetworkingchallenge.

I think I know what the issue is: "--ckpt-path" should point to the specific checkpoint, not the folder it is in. Try using "--ckpt-path ckpt/Baseline_cbr_mb/best_08_10/22-41.9264" instead. Let me know if it works or the issue still persists.

from gnnetworkingchallenge.

Yes, it has worked. Thanks !

from gnnetworkingchallenge.

You're welcome, I'm glad it was solved!

@jonyejin Can you try to see if it works for you now?

from gnnetworkingchallenge.

It worked out for me too! Thanks to @CarlosGuemS and @gblessed. I'll close the issue!

from gnnetworkingchallenge.