I completed the first exercise successfully, but things stopped working regarding exercises two and three. I got an error of "
AttributeError: The layer “lstm_4” has multiple inbound nodes, with different output shapes. Hence the notion of “output shape” is ill-defined for the layer. Use get_output_shape_at(node_index) instead." even though the number of total and trainable parameters is the same in the summary (picture below). I tried to restart the kernel and clear the output, but this didn’t help (after restarting it showed All tests passed for exercise one only, but it didn’t count in the grader), a similar thing happened with exercise 3. When I tried to submit my grades it gave me 0/100.
The “multiple inbound nodes” error can have several different causes.
The first thing to be aware of is that in this assignment, every time you modify anything in the model, you must restart the kernel and run all of the cells again.
This is because the model is a global variable, and any changes in any cell make the previous results invalid.
So that’s the first thing to try.
I don’t know man. I feel like I tried everything. I restarted and removed the outputs of the kernel many times, nothing worked.
Ok, it’s probably time to go to “plan B” and actually look at your notebook. Please see your DMs for a message from me.
To close the loop on the public thread, there was a bug in music_inference_model where the x value for the next timestep was not being computed correctly.