CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutSign UpSign In
amanchadha

Real-time collaboration for Jupyter Notebooks, Linux Terminals, LaTeX, VS Code, R IDE, and more,
all in one place. Commercial Alternative to JupyterHub.

GitHub Repository: amanchadha/coursera-deep-learning-specialization
 Path: blob/master/C4 - Convolutional Neural Networks/Week 2/ResNets/public_tests.py
Views: 4818
1
from termcolor import colored

2
import tensorflow as tf

3
from tensorflow.keras.initializers import random_uniform, glorot_uniform, constant, identity

4
import numpy as np

5


6
def identity_block_test(target):

7
    np.random.seed(1)

8
    #X = np.random.randn(3, 4, 4, 6).astype(np.float32)

9
    X1 = np.ones((1, 4, 4, 3)) * -1

10
    X2 = np.ones((1, 4, 4, 3)) * 1

11
    X3 = np.ones((1, 4, 4, 3)) * 3

12


13
    X = np.concatenate((X1, X2, X3), axis = 0).astype(np.float32)

14


15
    A3 = target(X,

16
                f = 2,

17
                filters = [4, 4, 3],

18
                initializer=lambda seed=0:constant(value=1),

19
                training=False)

20


21


22
    A3np = A3.numpy()

23
    assert tuple(A3np.shape) == (3, 4, 4, 3), "Shapes does not match. This is really weird"

24
    assert np.all(A3np >= 0), "The ReLu activation at the last layer is missing"

25
    resume = A3np[:,(0,-1),:,:].mean(axis = 3)

26


27
    assert np.floor(resume[1, 0, 0]) == 2 * np.floor(resume[1, 0, 3]), "Check the padding and strides"

28
    assert np.floor(resume[1, 0, 3]) == np.floor(resume[1, 1, 0]),     "Check the padding and strides"

29
    assert np.floor(resume[1, 1, 0]) == 2 * np.floor(resume[1, 1, 3]), "Check the padding and strides"

30
    assert np.floor(resume[1, 1, 0]) == 2 * np.floor(resume[1, 1, 3]), "Check the padding and strides"

31


32
    assert resume[1, 1, 0] - np.floor(resume[1, 1, 0]) > 0.7, "Looks like the BatchNormalization units are not working"

33


34
    assert np.allclose(resume, 

35
                       np.array([[[0.0,       0.0,       0.0,        0.0],

36
                                  [0.0,       0.0,       0.0,        0.0]],

37
                                 [[192.71236, 192.71236, 192.71236,  96.85619],

38
                                  [ 96.85619,  96.85619,  96.85619,  48.9281 ]],

39
                                 [[578.1371,   578.1371,  578.1371,  290.56854],

40
                                  [290.56854,  290.56854, 290.56854, 146.78427]]]), atol = 1e-5 ), "Wrong values with training=False"

41
    

42
    np.random.seed(1)

43
    A4 = target(X,

44
                f = 3,

45
                filters = [3, 3, 3],

46
                initializer=lambda seed=7:constant(value=1),

47
                training=True)

48
    A4np = A4.numpy()

49
    resume = A4np[:,(0,-1),:,:].mean(axis = 3)

50
    assert np.allclose(resume, 

51
                         np.array([[[0.,         0.,        0.,      0.,        ],

52
                                  [0.,         0.,        0.,        0.,        ]],

53
                                 [[0.37394285, 0.37394285, 0.37394285, 0.37394285],

54
                                  [0.37394285, 0.37394285, 0.37394285, 0.37394285]],

55
                                 [[3.2379014,  4.1394243,  4.1394243,  3.2379014 ],

56
                                  [3.2379014,  4.1394243,  4.1394243,  3.2379014 ]]]), atol = 1e-5 ), "Wrong values with training=True"

57


58
    print(colored("All tests passed!", "green"))

59


60