Errors when testing the train.py with tensorflow as backend

[zhangyu2ustc]

Hi, I was trying to test the keras-gcn package with the cora dataset. But there is always an error when I try to run the command: H = GraphConvolution(16, support, activation='relu', kernel_regularizer=l2(5e-4))([H]+G)

Output message: Loading cora dataset... Dataset has 2708 nodes, 5429 edges, 1433 features.

InvalidArgumentError: Dimensions must be equal, but are 4299 and 1433 for 'graph_convolution_2/MatMul' (op: 'MatMul') with input shapes: [?,4299], [1433,16].

The dataset has a shape of (2708, 1433) for input and (2708, 7) for output.

Do you have a clue what causes this error and how can I fix it ? My keras version is 2.1.1, with tensorflow (version 1.4.0) as backend.

Thanks! Best, Yu.

[tkipf]

Interesting- thanks for reporting. Did you make any changes to the code?

[zhangyu2ustc]

Thanks for the quick reply.

No, I did not make changes to the script, but the codes into jupyter-notebook for running. Is that a problem?

[tkipf]

Can you try running the script directly from the command line?

[zhangyu2ustc]

Sorry, I have errors during installation of the package: error: [Errno 5] Input/output error: '${HOME}/tensorflow/lib/python3.5/site-packages/kegra-0.0.1-py3.5.egg

This is probably due to the permission issues in our local clusters. That's why I want to test it using jupyter. Do you have any idea what might cause the problem of changing the dimensionality: [?,4299]

[tkipf]

I'm not sure how I can reproduce your error. But instead of installing the package you can also add the folder containing it to your PYTHONPATH environment variable, e.g. by running

export PYTHONPATH="~/code/keras-gcn/:$PYTHONPATH"

Simply replace ~/code/keras-gcn/ with the path where your package is located.

After the folder is added to your PYTHONPATH environment variable, you can run the model from your command line (without installing the package).

[zhangyu2ustc]

Thanks! It worked in this way!

[zhangyu2ustc]

The error message showed up again when I change the FILTER method from 'localpool' to 'chebyshev' even by running "python train.py" from terminal

ValueError: Dimensions must be equal, but are 4299 and 1433 for 'graph_convolution_1/MatMul' (op: 'MatMul') with input shapes: [?,4299], [1433,16].

[tkipf]

Ah, thanks for catching this - this is something I can look into.

[tkipf]

Fixed with https://github.com/tkipf/keras-gcn/commit/eb89564a0e865640c11283991685d80c84bc602a