Unit test failure in named_image_VGG19_test.py

[iandow]

I'm running the python unit tests on Ubuntu 16.04.3 LTS Xenial on x86 hardware with the following command:

$ SPARK_HOME=/home/mapr/development/spark-2.2.1-bin-hadoop2.7 PYSPARK_PYTHON=python3 SCALA_VERSION=2.11.8 SPARK_VERSION=2.1.1 ./python/run-tests.sh

After a few minutes, I'm getting the following failure:

Using Spark's default log4j profile: org/apache/spark/log4j-defaults.properties
Setting default log level to "WARN".
To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use setLogLevel(newLevel).
17/12/26 13:15:39 WARN NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Using TensorFlow backend.
[Stage 1:>                                                          (0 + 8) / 8]Using TensorFlow backend.
Using TensorFlow backend.
Using TensorFlow backend.
Using TensorFlow backend.
Using TensorFlow backend.
Using TensorFlow backend.
Using TensorFlow backend.
Using TensorFlow backend.
[Stage 1:===================================================>       (7 + 1) / 8]Using TensorFlow backend.
ok
Run sparkDL predictor on manually-resized images and compare result to the ... 17/12/26 13:20:44 WARN TaskSetManager: Stage 6 contains a task of very large size (740 KB). The maximum recommended task size is 100 KB.
17/12/26 13:20:48 WARN TaskSetManager: Stage 7 contains a task of very large size (740 KB). The maximum recommended task size is 100 KB.
ok
" ... FAIL
Tests that featurizer returns (almost) the same values as Keras. ... ok
Tests that featurizer fits into an MLlib Pipeline. ... 17/12/26 13:21:13 WARN BLAS: Failed to load implementation from: com.github.fommil.netlib.NativeSystemBLAS
17/12/26 13:21:13 WARN BLAS: Failed to load implementation from: com.github.fommil.netlib.NativeRefBLAS
ok
Tests that predictor with decoded=true returns reasonable values. ... ok

======================================================================
FAIL: "
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/mapr/development/spark-deep-learning/python/tests/transformers/named_image_test.py", line 127, in test_buildtfgraphforname
    np.testing.assert_array_almost_equal(kerasPredict, tfPredict)
  File "/usr/local/lib/python3.5/dist-packages/numpy/testing/utils.py", line 962, in assert_array_almost_equal
    precision=decimal)
  File "/usr/local/lib/python3.5/dist-packages/numpy/testing/utils.py", line 778, in assert_array_compare
    raise AssertionError(msg)
AssertionError:
Arrays are not almost equal to 6 decimals

(mismatch 0.01999999999999602%)
 x: array([[  7.134266e-07,   2.800847e-06,   1.959372e-06, ...,
          3.629200e-06,   2.826131e-05,   2.570361e-04],
       [  1.193586e-04,   2.015762e-02,   2.432920e-06, ...,...
 y: array([[  7.134280e-07,   2.800855e-06,   1.959379e-06, ...,
          3.629210e-06,   2.826137e-05,   2.570372e-04],
       [  1.193585e-04,   2.015762e-02,   2.432913e-06, ...,...
-------------------- >> begin captured stdout << ---------------------
Converted 38 variables to const ops.

--------------------- >> end captured stdout << ----------------------