Help with C5W1 Jazz Improv with LSTM assignment

Course Q&A Deep Learning Specialization Sequence Models
1

Link to assignment: Coursera | Online Courses & Credentials From Top Educators. Join for Free | Coursera

I need help with this assignment and I have already looked at previous posts and tried their suggestions.
In the unit test of after the definition of djmodel, I am getting an error which I think is due to the fact that the test does not expect the slice operation to appear as a layer in the model summary. I initially implemented the slice, as suggested in the notebook, using the python array index syntax, but I have also used tf.slice and tf.strided_slice and they also didn’t solve it. I have also played with changing method input and output params in all layers and it also didn’t help…
The error and the model summary are below:

Test failed at index 2 
 Expected value 

 ['TensorFlowOpLayer', [(None, 90)], 0] 

 does not match the input value: 

 ['Reshape', (None, 1, 90), 0]

Here is the output of the model summary:

Model: "model_1"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_2 (InputLayer)            [(None, 30, 90)]     0                                            
__________________________________________________________________________________________________
tf_op_layer_strided_slice_30 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_31 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_32 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_33 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_34 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_35 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_36 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_37 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_38 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_39 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_40 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_41 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_42 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_43 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_44 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_45 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_46 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_47 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_48 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_49 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_50 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_51 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_52 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_53 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_54 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_55 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_56 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_57 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_58 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
tf_op_layer_strided_slice_59 (T [(None, 90)]         0           input_2[0][0]                    
__________________________________________________________________________________________________
reshape (Reshape)               (None, 1, 90)        0           tf_op_layer_strided_slice_30[0][0
                                                                 tf_op_layer_strided_slice_31[0][0
                                                                 tf_op_layer_strided_slice_32[0][0
                                                                 tf_op_layer_strided_slice_33[0][0
                                                                 tf_op_layer_strided_slice_34[0][0
                                                                 tf_op_layer_strided_slice_35[0][0
                                                                 tf_op_layer_strided_slice_36[0][0
                                                                 tf_op_layer_strided_slice_37[0][0
                                                                 tf_op_layer_strided_slice_38[0][0
                                                                 tf_op_layer_strided_slice_39[0][0
                                                                 tf_op_layer_strided_slice_40[0][0
                                                                 tf_op_layer_strided_slice_41[0][0
                                                                 tf_op_layer_strided_slice_42[0][0
                                                                 tf_op_layer_strided_slice_43[0][0
                                                                 tf_op_layer_strided_slice_44[0][0
                                                                 tf_op_layer_strided_slice_45[0][0
                                                                 tf_op_layer_strided_slice_46[0][0
                                                                 tf_op_layer_strided_slice_47[0][0
                                                                 tf_op_layer_strided_slice_48[0][0
                                                                 tf_op_layer_strided_slice_49[0][0
                                                                 tf_op_layer_strided_slice_50[0][0
                                                                 tf_op_layer_strided_slice_51[0][0
                                                                 tf_op_layer_strided_slice_52[0][0
                                                                 tf_op_layer_strided_slice_53[0][0
                                                                 tf_op_layer_strided_slice_54[0][0
                                                                 tf_op_layer_strided_slice_55[0][0
                                                                 tf_op_layer_strided_slice_56[0][0
                                                                 tf_op_layer_strided_slice_57[0][0
                                                                 tf_op_layer_strided_slice_58[0][0
                                                                 tf_op_layer_strided_slice_59[0][0
__________________________________________________________________________________________________
a0 (InputLayer)                 [(None, 64)]         0                                            
__________________________________________________________________________________________________
c0 (InputLayer)                 [(None, 64)]         0                                            
__________________________________________________________________________________________________
lstm (LSTM)                     [(None, 64), (None,  39680       reshape[30][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[31][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[32][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[33][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[34][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[35][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[36][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[37][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[38][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[39][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[40][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[41][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[42][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[43][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[44][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[45][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[46][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[47][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[48][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[49][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[50][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[51][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[52][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[53][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[54][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[55][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[56][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[57][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[58][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
                                                                 reshape[59][0]                   
                                                                 a0[0][0]                         
                                                                 c0[0][0]                         
__________________________________________________________________________________________________
dense (Dense)                   (None, 90)           5850        lstm[30][1]                      
                                                                 lstm[31][1]                      
                                                                 lstm[32][1]                      
                                                                 lstm[33][1]                      
                                                                 lstm[34][1]                      
                                                                 lstm[35][1]                      
                                                                 lstm[36][1]                      
                                                                 lstm[37][1]                      
                                                                 lstm[38][1]                      
                                                                 lstm[39][1]                      
                                                                 lstm[40][1]                      
                                                                 lstm[41][1]                      
                                                                 lstm[42][1]                      
                                                                 lstm[43][1]                      
                                                                 lstm[44][1]                      
                                                                 lstm[45][1]                      
                                                                 lstm[46][1]                      
                                                                 lstm[47][1]                      
                                                                 lstm[48][1]                      
                                                                 lstm[49][1]                      
                                                                 lstm[50][1]                      
                                                                 lstm[51][1]                      
                                                                 lstm[52][1]                      
                                                                 lstm[53][1]                      
                                                                 lstm[54][1]                      
                                                                 lstm[55][1]                      
                                                                 lstm[56][1]                      
                                                                 lstm[57][1]                      
                                                                 lstm[58][1]                      
                                                                 lstm[59][1]                      
==================================================================================================
Total params: 45,530
Trainable params: 45,530
Non-trainable params: 0
2

It’s a little hard to read the output of the model summary as they show it. For clarity, I added the following cell to my notebook:

print("Generated model:")
for index, a in enumerate(summary(model)):
    print(f"layer {index}: {a}")
print("Expected model:")
for index, a in enumerate(djmodel_out):
    print(f"layer {index}: {a}")

Here’s the first relevant part of the output I get with code that passes all the tests and the grader:

Generated model:
layer 0: ['InputLayer', [(None, 30, 90)], 0]
layer 1: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 2: ['Reshape', (None, 1, 90), 0]
layer 3: ['InputLayer', [(None, 64)], 0]
layer 4: ['InputLayer', [(None, 64)], 0]
layer 5: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 6: ['LSTM', [(None, 64), (None, 64), (None, 64)], 39680, [(None, 1, 90), (None, 64), (None, 64)], 'tanh']
layer 7: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 8: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 9: ['TensorFlowOpLayer', [(None, 90)], 0]

What do you see when you try that? You must have ordered the calls differently in your code to produce a different “compute graph”.

3

Also note that it is not necessary to use tf.slice in order to perform the slicing: they give you some hints in the instructions about how to do it just by indexing the array on the second (timestep) dimension.

4

Hi!
Thanks for the reply.
I did the slicing with the python array slice syntax (based on the hint) and I was just mentioning other methods I tried in my post.
Here’s what I get from running your code segment:

Generated model:
layer 0: ['InputLayer', [(None, 30, 90)], 0]
layer 1: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 2: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 3: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 4: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 5: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 6: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 7: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 8: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 9: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 10: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 11: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 12: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 13: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 14: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 15: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 16: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 17: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 18: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 19: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 20: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 21: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 22: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 23: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 24: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 25: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 26: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 27: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 28: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 29: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 30: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 31: ['Reshape', (None, 1, 90), 0]
layer 32: ['InputLayer', [(None, 64)], 0]
layer 33: ['InputLayer', [(None, 64)], 0]
layer 34: ['LSTM', [(None, 64), (None, 64), (None, 64)], 39680, [(None, 1, 90), (None, 64), (None, 64)], 'tanh']
layer 35: ['Dense', (None, 90), 5850, 'softmax']
Expected model:
layer 0: ['InputLayer', [(None, 30, 90)], 0]
layer 1: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 2: ['Reshape', (None, 1, 90), 0]
layer 3: ['InputLayer', [(None, 64)], 0]
layer 4: ['InputLayer', [(None, 64)], 0]
layer 5: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 6: ['LSTM', [(None, 64), (None, 64), (None, 64)], 39680, [(None, 1, 90), (None, 64), (None, 64)], 'tanh']
layer 7: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 8: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 9: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 10: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 11: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 12: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 13: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 14: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 15: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 16: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 17: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 18: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 19: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 20: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 21: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 22: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 23: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 24: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 25: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 26: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 27: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 28: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 29: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 30: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 31: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 32: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 33: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 34: ['TensorFlowOpLayer', [(None, 90)], 0]
layer 35: ['Dense', (None, 90), 5850, 'softmax']

It seems like mine is out of order for some reason. I didn’t change the order of ops from its initial ordering in the notebook, nor do I see any other logical ordering for them from what they are.

5

But your compute graph is clearly very different than the expected one, so you must be missing something in terms of translating the instructions into code.

I realize that may not be enough of a hint to be much help. Maybe it’s time to just look at your code. We can’t do that on a public thread, but please check your DMs for a message from me about how to proceed.

6

To close the loop on the public thread, the bug was one that I’d seen before, but I had unfortunately forgotten what effect that has on the compute graph. The problem was on the LSTM invocation in the for loop: the initial state was the same for every timestep (a0, c0). But that’s not how RNNs work: the cell state at the start of each timestep is the cell state that was output by the previous timestep, right?

1 Like