Merge pull request #98 from bartvm/dataset_tests

Fixes MNIST and runs test on Travis

.travis.yml

@@ -12,29 +12,51 @@ env:
 before_install:
   - |
       if [ $TESTS ]; then
+        # Setup Python environment with BLAS libraries
         sudo apt-get install -qq libatlas3gf-base libatlas-dev liblapack-dev gfortran
         wget -q http://repo.continuum.io/miniconda/Miniconda-latest-Linux-x86_64.sh -O miniconda.sh
         chmod +x miniconda.sh
         ./miniconda.sh -b
         export PATH=/home/travis/miniconda/bin:$PATH
         conda update -q --yes conda
+        # Download MNIST for tests
+        (mkdir mnist
+        cd mnist
+        curl -O http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz \
+             -O http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz \
+             -O http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz \
+             -O http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
+        gunzip *-ubyte.gz)
+        export BLOCKS_DATA_PATH=$PWD
       fi
 install:
+  # Install all Python dependencies
   - |
       if [ $TESTS ]; then
         conda install -q --yes python=$TRAVIS_PYTHON_VERSION nose numpy pip coverage six scipy
         pip install -q --no-deps git+git://github.com/Theano/Theano.git
         pip install -q nose2[coverage-plugin] coveralls
         git clone -q git://github.com/lisa-lab/pylearn2.git
-        (cd pylearn2; python setup.py -q develop)
+        (cd pylearn2
+        python setup.py -q develop)
       fi
   - |
       if [ $FORMAT ]; then
         pip install -q flake8
         pip install -q git+git://github.com/bartvm/pep257.git@numpy
       fi
 script:
-  - if [ $TESTS ]; then THEANO_FLAGS=floatX=$TESTS,blas.ldflags='-lblas -lgfortran' coverage run --source=blocks -m nose2.__main__ tests; fi
-  - if [ $FORMAT ]; then flake8 blocks tests; pep257 blocks --numpy --ignore=D100,D101,D102,D103; fi
+  - |
+      if [ $TESTS ]; then
+        THEANO_FLAGS=floatX=$TESTS,blas.ldflags='-lblas -lgfortran' \
+        # Running nose2 within coverage makes imports count towards coverage
+        coverage run --source=blocks -m nose2.__main__ tests
+      fi
+  - |
+      if [ $FORMAT ]; then
+        flake8 blocks tests
+        # Ignore D100-103 errors (non-existing docstrings)
+        pep257 blocks --numpy --ignore=D100,D101,D102,D103
+      fi
 after_script:
   - if [ $TESTS ]; then coveralls; fi

blocks/datasets/__init__.py

@@ -44,7 +44,6 @@ def __init__(self, sources=None):
                 raise ValueError("Unable to provide requested sources")
             self.sources = sources
 
-    @abstractmethod
     def open(self):
         """Return the state if the dataset requires one.
 

blocks/datasets/mnist.py

@@ -40,11 +40,15 @@ class MNIST(Dataset):
         The first example to load
     stop : int, optional
         The last example to load
+    binary : bool, optional
+        If ``True``, returns binary (black/white) images instead of
+        grayscale. ``False`` by default.
 
     """
     sources = ('features', 'targets')
 
-    def __init__(self, which_set, start=None, stop=None, **kwargs):
+    def __init__(self, which_set, start=None, stop=None, binary=False,
+                 **kwargs):
         if which_set == 'train':
             data = 'train-images-idx3-ubyte'
             labels = 'train-labels-idx1-ubyte'
@@ -53,21 +57,21 @@ def __init__(self, which_set, start=None, stop=None, **kwargs):
             labels = 't10k-labels-idx1-ubyte'
         else:
             raise ValueError("MNIST only has a train and test set")
+        data_path = os.path.join(config.data_path, 'mnist')
         X = read_mnist_images(
-            os.path.join(config.data_path, data),
-            theano.config.floatX)[start:stop]
+            os.path.join(data_path, data),
+            'bool' if binary else theano.config.floatX)[start:stop]
         X = X.reshape((X.shape[0], numpy.prod(X.shape[1:])))
         y = read_mnist_labels(
-            os.path.join(config.data_path, labels))[start:stop, numpy.newaxis]
+            os.path.join(data_path, labels))[start:stop, numpy.newaxis]
         self.X, self.y = X, y
         self.num_examples = len(X)
         self.default_scheme = SequentialScheme(self.num_examples, 1)
         super(MNIST, self).__init__(**kwargs)
 
-    def get_data(self, state=None, request=None, sources=None):
-        data = dict(zip(self.sources, (self.X, self.y)))
-        sources = self.sources if sources is None else sources
-        return tuple(data[source][request] for source in sources)
+    def get_data(self, request=None):
+        data = dict(zip(('features', 'targets'), (self.X, self.y)))
+        return tuple(data[source][request] for source in self.sources)
 
 
 def read_mnist_images(filename, dtype=None):

tests/datasets/test_mnist.py

@@ -0,0 +1,24 @@
+from numpy.testing import assert_raises
+
+from blocks.datasets.mnist import MNIST
+
+
+def test_mnist():
+    mnist_train = MNIST('train', start=20000)
+    assert len(mnist_train.X) == 40000
+    assert len(mnist_train.y) == 40000
+    mnist_test = MNIST('test', sources=('targets',))
+    assert len(mnist_test.X) == 10000
+    assert len(mnist_test.y) == 10000
+
+    first_feature, first_target = mnist_train.get_data(request=[0])
+    assert first_feature.shape == (1, 784)
+    assert first_target.shape == (1, 1)
+
+    first_target, = mnist_test.get_data(request=[0, 1])
+    assert first_target.shape == (2, 1)
+
+    binary_mnist = MNIST('test', binary=True, sources=('features',))
+    first_feature, = binary_mnist.get_data(request=[0])
+    assert first_feature.dtype.kind == 'b'
+    assert_raises(ValueError, MNIST, 'valid')