Error running train.py about keras-gcn HOT 8 CLOSED

Thanks for reporting! Which version of keras are you using?

from keras-gcn.

2.0.5, I just saw from the other issue that this version is not supported yet! Thanks, I'll wait for the fix =)

from keras-gcn.

I'll look into it soon - will keep you updated!

from keras-gcn.

I am also facing the same issue using Keras 2.0.4.

from keras-gcn.

I upgraded gcn for keras 2.0.5, you can try it in https://github.com/buaahsh/keras-gcn.
The code needs to be modified in three place in order to handle keras 2.X.

• initializations -> initializers

• add self.is_placeholder = True in GraphInputLayer

• change get_output_shape_for into compute_output_shape in GraphConvolution

from keras-gcn.

Thanks! Feel free to make a pull request

from keras-gcn.

My bad, I forgot to re-install. Its working perfectly.

Hi @buaahsh
I downloaded your repo. But it still shows the same error on line 57 of train.py

ValueError: Dimensions must be equal, but are 16 and 1432 for 'graph_convolution_4/MatMul' (op: 'MatMul') with input shapes: [?,16], [1432,7].

from keras-gcn.

Fixed with #10

from keras-gcn.