My first exercise passes all tests. In the second exercise, I am getting the error: Graph disconnected: cannot obtain value for tensor Tensor(“input_7:0”, shape=(None, 1, 90), dtype=float32) at layer “lstm_1”. The following previous layers were accessed without issue:

My ‘out’ is the same as in the working first exercise, so I assume it is correct. It has shape (None, 90) for the first iteration, and has shape (90,90) for subsequent iterations. After applying argmax to out, x has shape (90,) and after applying one_hot to x (second argument, n_values) x has shape (90,90) and after applying RepeatVector(1) to x, x has shape (90,1,90).

I have restarted the kernel several times, rerun all upstream cells, etc., no change in behavior.

Any hint what I’m doing wrong would be most appreciated! Thanks!

FYI, I’ve also tried -1 as the axis parameter in argmax. When I do that, my dimensions are:

after argmax: (None,)

after one_hot: (None, 90)

after RepeatVector: (None, 1, 90)

But I still get the same “Graph Disconnected” error, still no clue how to proceed! Thanks in advance for your help!