Exercise 3 - compute_layer_style_cost

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

Hi
In this exercise in addition to change the dimension, we should transpose the dimension. I did it but it didn’t work?
What is my problem? or how should I fix it?
a_S = tf.transpose(tf.reshape(a_S,shape=[m,n_Hn_W,n_C]),perm=[0,1 ,2])
a_G = tf.transpose(tf.reshape(a_G,shape=[m,n_Hn_W,n_C]),perm=[0,1 ,2])

2

What’s the error you receive?

3

InvalidArgumentError Traceback (most recent call last)
in
2 a_S = tf.random.normal([1, 4, 4, 3], mean=1, stddev=4)
3 a_G = tf.random.normal([1, 4, 4, 3], mean=1, stddev=4)
----> 4 J_style_layer_GG = compute_layer_style_cost(a_G, a_G)
5 J_style_layer_SG = compute_layer_style_cost(a_S, a_G)
6

in compute_layer_style_cost(a_S, a_G)
25
26 # Computing gram_matrices for both images S and G (≈2 lines)
—> 27 GS = gram_matrix(a_S)
28 GG = gram_matrix(a_G)
29

in gram_matrix(A)
13 #(≈1 line)
14
—> 15 GA = tf.linalg.matmul(A,tf.transpose(A))
16
17 ### END CODE HERE

/usr/local/lib/python3.6/dist-packages/tensorflow/python/util/dispatch.py in wrapper(*args, **kwargs)
199 “”“Call target, and fall back on dispatchers if there is a TypeError.”""
200 try:
→ 201 return target(*args, **kwargs)
202 except (TypeError, ValueError):
203 # Note: convert_to_eager_tensor currently raises a ValueError, not a

/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py in matmul(a, b, transpose_a, transpose_b, adjoint_a, adjoint_b, a_is_sparse, b_is_sparse, name)
3215 adjoint_b = True
3216 return gen_math_ops.batch_mat_mul_v2(
→ 3217 a, b, adj_x=adjoint_a, adj_y=adjoint_b, name=name)
3218
3219 # Neither matmul nor sparse_matmul support adjoint, so we conjugate

/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/gen_math_ops.py in batch_mat_mul_v2(x, y, adj_x, adj_y, name)
1544 return _result
1545 except _core._NotOkStatusException as e:
→ 1546 _ops.raise_from_not_ok_status(e, name)
1547 except _core._FallbackException:
1548 pass

/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/ops.py in raise_from_not_ok_status(e, name)
6841 message = e.message + (" name: " + name if name is not None else “”)
6842 # pylint: disable=protected-access
→ 6843 six.raise_from(core._status_to_exception(e.code, message), None)
6844 # pylint: enable=protected-access
6845

/usr/local/lib/python3.6/dist-packages/six.py in raise_from(value, from_value)

InvalidArgumentError: In[0] mismatch In[1] shape: 1 vs. 3: [16,3,1] [1,3,16] 0 0 [Op:BatchMatMulV2]

4

I am stuck on the same question. I know the problem is that we’re not reshaping correctly. Look at your error:

InvalidArgumentError: In[0] mismatch In[1] shape: 1 vs. 3: [16,3,1] [1,3,16] 0 0 [Op:BatchMatMulV2]

How can matmul do this calculation? I believe it needs to be unrolled further. It should become [16, 3x1] [3x1, 16]. In other words, 2 dimensional. But I can’t solve this question either. I have been stuck on it for so long

5

Look at this old thread: Course 4, week 4, coding assignment 2, compute_layer_style_cost

6

You have 3 dimensions in the output of the reshape. That’s not what the instructions tell you to do. When in doubt, it never hurts to read the instructions again more carefully. :nerd_face: Also note that it is not necessary to use the permute argument in this case, but it is critical that you use both reshape and transpose: you can’t just directly reshape to the final shape that you want because it ends up scrambling the data. See the link that @crackingthecode gave just above for more explanation.