Error at travis CI

[ita9naiwa]

1. Cannot find approximate_als_test
   in travis
   https://travis-ci.org/benfred/implicit/jobs/543107113#L740

    return self.loadTestsFromModule(obj)
  File "/home/travis/virtualenv/python2.7.14/lib/python2.7/site-packages/setuptools/command/test.py", line 52, in loadTestsFromModule
    tests.append(self.loadTestsFromName(submodule))
  File "/opt/python/2.7.14/lib/python2.7/unittest/loader.py", line 100, in loadTestsFromName
    parent, obj = obj, getattr(obj, part)
AttributeError: 'module' object has no attribute 'approximate_als_test'
The command "$PY setup.py test" exited with 1.
cache.2
store build cache

I think this is because the test environment have no annoy, nmslib, and faiss

try:
    import annoy  # noqa

    class AnnoyALSTest(unittest.TestCase, TestRecommenderBaseMixin):
        def _get_model(self):
            return AnnoyAlternatingLeastSquares(factors=2, regularization=0, use_gpu=False)
except ImportError:
    pass

try:
    import nmslib  # noqa

    class NMSLibALSTest(unittest.TestCase, TestRecommenderBaseMixin):
        def _get_model(self):
            return NMSLibAlternatingLeastSquares(factors=2, regularization=0,
                                                 index_params={'post': 2}, use_gpu=False)
except ImportError:
    pass

try:
    import faiss  # noqa

    class FaissALSTest(unittest.TestCase, TestRecommenderBaseMixin):
        def _get_model(self):
            return FaissAlternatingLeastSquares(nlist=1, nprobe=1, factors=2, regularization=0,
                                                use_gpu=False)

    if HAS_CUDA:
        class FaissALSGPUTest(unittest.TestCase, TestRecommenderBaseMixin):
            __regularization = 0

            def _get_model(self):
                return FaissAlternatingLeastSquares(nlist=1, nprobe=1, factors=32,
                                                    regularization=self.__regularization,
                                                    use_gpu=True)

            def test_similar_items(self):
                # For the GPU version, we currently have to have factors be a multiple of 32
                # (limitation that I think is caused by how we are currently calculating the
                # dot product in CUDA, TODO: eventually should fix that code).
                # this causes the test_similar_items call to fail if we set regularization to 0
                self.__regularization = 1.0
                try:
                    super(FaissALSGPUTest, self).test_similar_items()
                finally:
                    self.__regularization = 0.0

            def test_large_recommend(self):
                # the GPU version of FAISS can't return more than 1K result (and will assert/exit)
                # this tests out that we fall back in this case to the exact version and don't die
                plays = self.get_checker_board(2048)
                model = self._get_model()
                model.fit(plays, show_progress=False)

                recs = model.similar_items(0, N=1050)
                self.assertEqual(recs[0][0], 0)

                recs = model.recommend(0, plays.T.tocsr(), N=1050)
                self.assertEqual(recs[0][0], 0)

except ImportError:
    pass

I circumbented this by adding the unittest.TestCase subclass that do nothing.

class DoNothingTest(unittest.TestCase):
    # it's because build error
    def test_nothing(self):
        pass

How can we do better to solve this problem...?

[ita9naiwa]

2. isort failed

$ isort -c **/*.py
ERROR: /home/travis/build/benfred/implicit/benchmarks/benchmark_als.py Imports are incorrectly sorted.
ERROR: /home/travis/build/benfred/implicit/benchmarks/benchmark_spark.py Imports are incorrectly sorted.
The command "isort -c **/*.py" exited with 1.

It works correctly when tested my local PC, but failed.
I think it seems the version of isort.

[ita9naiwa]

3. Re-installation of Numpy

#180 Solved.

[ita9naiwa]

2 solved.
#217

[ita9naiwa]

1. solved
   Fix Travis build Error #219