C4W4 assignment dimension issues

Course Q&A TensorFlow Developer Professional Sequences, Time Series and Prediction
1

I’m in the cell trying to test my uncompiled model and am getting various dimension related errors.

If I dumb down the model to just a single Dense neuron with input shape[None,G.WINDOW_SIZE] I get the most helpful error:

WARNING:tensorflow:Model was constructed with shape (None, None, 64) for input KerasTensor(type_spec=TensorSpec(shape=(None, None, 64), dtype=tf.float32, name=‘dense_56_input’), name=‘dense_56_input’, description=“created by layer ‘dense_56_input’”), but it was called on an input with incompatible shape (None, None).

I’m surprised by the actual train_set dimension of None,None, but my CSV reader seems to be working and the rest of the data setup is provided for us.

I’m stuck… any hints based on above?

Thanks,

T.

2

So… it looks like this is environmental. My model works fine on the coursera server, but won’t run on my local installation that’s run everything else.

Any guesses on what might be missing and why the uncompiled predict won’t run on my instance?

T.

3

Please print(model.summary()) to understand if the model output shape matches the dataset output shape.