Fix flake8 in tests

blocks/bricks/__init__.py

@@ -7,8 +7,8 @@
 from .interfaces import (Activation, Feedforward, Initializable, LinearLike,
                          Random)
 from .recurrent import (BaseRecurrent, SimpleRecurrent, LSTM, GatedRecurrent,
-                        Bidirectional, RecurrentStack, RECURRENTSTACK_SEPARATOR,
-                        recurrent)
+                        Bidirectional, RecurrentStack,
+                        RECURRENTSTACK_SEPARATOR, recurrent)
 from .simple import (Linear, Bias, Maxout, LinearMaxout, Identity, Tanh,
                      Logistic, Softplus, Rectifier, LeakyRectifier,
                      Softmax, NDimensionalSoftmax)

blocks/bricks/base.py

@@ -32,6 +32,7 @@ def create_unbound_method(func, cls):
     if six.PY3:
         return func
 
+
 # Rename built-in property to avoid conflict with Application.property
 property_ = property
 
@@ -806,6 +807,7 @@ def __bool__(self):
 
     __nonzero__ = __bool__
 
+
 NoneAllocation = LazyNone('NoneAllocation')
 NoneInitialization = LazyNone('NoneInitialization')
 

blocks/bricks/recurrent/__init__.py

@@ -1,3 +1,8 @@
 from .base import BaseRecurrent, recurrent
 from .architectures import SimpleRecurrent, LSTM, GatedRecurrent
 from .misc import Bidirectional, RecurrentStack, RECURRENTSTACK_SEPARATOR
+
+
+__all__ = ("BaseRecurrent", "recurrent", "SimpleRecurrent", "LSTM",
+           "GatedRecurrent", "Bidirectional", "RecurrentStack",
+           "RECURRENTSTACK_SEPARATOR")

blocks/bricks/recurrent/misc.py

@@ -58,6 +58,7 @@ def get_dim(self, name):
             return self.prototype.get_dim(name) * 2
         return self.prototype.get_dim(name)
 
+
 RECURRENTSTACK_SEPARATOR = '#'
 
 

blocks/monitoring/aggregation.py

@@ -217,6 +217,7 @@ def get_aggregator(self):
         return self._build_aggregator(tensor.minimum(self.storage,
                                                      self.variable))
 
+
 minimum = partial(_simple_aggregation, Minimum)
 
 
@@ -227,6 +228,7 @@ def get_aggregator(self):
         return self._build_aggregator(tensor.maximum(self.storage,
                                                      self.variable))
 
+
 maximum = partial(_simple_aggregation, Maximum)
 
 
@@ -251,6 +253,7 @@ def get_aggregator(self):
         return self._build_aggregator(tensor.concatenate([self.storage,
                                                           self.variable]))
 
+
 concatenate = partial(_simple_aggregation, Concatenate)
 
 

blocks/roles.py

@@ -75,146 +75,167 @@ def __repr__(self):
 class InputRole(VariableRole):
     pass
 
+
 #: The input of a :class:`~.bricks.Brick`
 INPUT = InputRole()
 
 
 class OutputRole(VariableRole):
     pass
 
+
 #: The output of a :class:`~.bricks.Brick`
 OUTPUT = OutputRole()
 
 
 class CostRole(VariableRole):
     pass
 
+
 #: A scalar cost that can be used to train or regularize
 COST = CostRole()
 
 
 class PersistentRole(VariableRole):
     pass
 
+
 # Any persistent quantity that should be saved as part of the model
 PERSISTENT = PersistentRole()
 
 
 class ParameterRole(PersistentRole):
     pass
 
+
 #: A parameter of the model
 PARAMETER = ParameterRole()
 
 
 class AuxiliaryRole(VariableRole):
     pass
 
+
 #: Variables added to the graph as annotations
 AUXILIARY = AuxiliaryRole()
 
 
 class WeightRole(ParameterRole):
     pass
 
+
 #: The weight matrices of linear transformations
 WEIGHT = WeightRole()
 
 
 class BiasRole(ParameterRole):
     pass
 
+
 #: Biases of linear transformations
 BIAS = BiasRole()
 
 
 class InitialStateRole(ParameterRole):
     pass
 
+
 #: Initial state of a recurrent network
 INITIAL_STATE = InitialStateRole()
 
 
 class FilterRole(WeightRole):
     pass
 
+
 #: The filters (kernels) of a convolution operation
 FILTER = FilterRole()
 
 
 class DropoutRole(VariableRole):
     pass
 
+
 #: Inputs with applied dropout
 DROPOUT = DropoutRole()
 
 
 class CollectedRole(VariableRole):
     pass
 
+
 #: The replacement of a variable collected into a single shared variable
 COLLECTED = CollectedRole()
 
 
 class CollectorRole(ParameterRole):
     pass
 
+
 #: A collection of parameters combined into a single shared variable
 COLLECTOR = CollectorRole()
 
 
 class AlgorithmStateRole(VariableRole):
     pass
 
+
 #: Shared variables used in algorithms updates
 ALGORITHM_STATE = AlgorithmStateRole()
 
 
 class AlgorithmHyperparameterRole(AlgorithmStateRole):
     pass
 
+
 #: hyperparameters accociated with algorithms
 ALGORITHM_HYPERPARAMETER = AlgorithmHyperparameterRole()
 
 
 class AlgorithmBufferRole(AlgorithmStateRole):
     pass
 
+
 #: buffers accociated with algorithms
 ALGORITHM_BUFFER = AlgorithmBufferRole()
 
 
 class BatchNormPopulationStatisticsRole(PersistentRole):
     pass
 
+
 #: base role for batch normalization population statistics
 BATCH_NORM_POPULATION_STATISTICS = BatchNormPopulationStatisticsRole()
 
 
 class BatchNormPopulationMeanRole(BatchNormPopulationStatisticsRole):
     pass
 
+
 #: mean activations accumulated over the dataset
 BATCH_NORM_POPULATION_MEAN = BatchNormPopulationMeanRole()
 
 
 class BatchNormPopulationStdevRole(BatchNormPopulationStatisticsRole):
     pass
 
+
 #: standard deviations of activations accumulated over the dataset
 BATCH_NORM_POPULATION_STDEV = BatchNormPopulationStdevRole()
 
 
 class BatchNormGraphVariableRole(VariableRole):
     pass
 
+
 #: base for roles used for within-graph batch normalization replacement
 BATCH_NORM_GRAPH_VARIABLE = BatchNormGraphVariableRole()
 
 
 class BatchNormOffsetRole(BatchNormGraphVariableRole):
     pass
 
+
 #: offset applied in a BatchNormalization application (or its
 #  batch-normalized replacement)
 BATCH_NORM_OFFSET = BatchNormOffsetRole()
@@ -223,6 +244,7 @@ class BatchNormOffsetRole(BatchNormGraphVariableRole):
 class BatchNormDivisorRole(BatchNormGraphVariableRole):
     pass
 
+
 #: divisor applied in a BatchNormalization application (or its
 #  batch-normalized replacement)
 BATCH_NORM_DIVISOR = BatchNormDivisorRole()
@@ -231,6 +253,7 @@ class BatchNormDivisorRole(BatchNormGraphVariableRole):
 class BatchNormMinibatchEstimateRole(BatchNormGraphVariableRole):
     pass
 
+
 #: role added to variables that are the result of a batch normalization
 #  replacement, rather than the original population statistics variables.
 BATCH_NORM_MINIBATCH_ESTIMATE = BatchNormMinibatchEstimateRole()
@@ -239,6 +262,7 @@ class BatchNormMinibatchEstimateRole(BatchNormGraphVariableRole):
 class BatchNormScaleParameterRole(ParameterRole):
     pass
 
+
 #: role given to the scale parameter, referred to as "scale" in the
 # batch normalization manuscript, applied after normalizing.
 BATCH_NORM_SCALE_PARAMETER = BatchNormScaleParameterRole()
@@ -247,6 +271,7 @@ class BatchNormScaleParameterRole(ParameterRole):
 class BatchNormShiftParameterRole(BiasRole):
     pass
 
+
 #: role given to the shift parameter, referred to as "beta" in the
 # batch normalization manuscript, applied after normalizing and scaling.
 # Inherits from BIAS, because there really is no functional difference

blocks/serialization.py

@@ -229,7 +229,7 @@ def secure_dump(object_, path, dump_function=dump, **kwargs):
         logger.debug("Moving the temporary file")
         shutil.move(temp.name, path)
         logger.debug("Dump finished")
-    except:
+    except:  # noqa: E722
         if "temp" in locals():
             os.remove(temp.name)
         raise
@@ -564,6 +564,7 @@ def _recreate_pygpu_array(context_name, content):
     context = theano.gpuarray.get_context(context_name)
     return pygpu.gpuarray.array(content, context=context)
 
+
 _ARRAY_TYPE_MAP = {numpy.ndarray: 'numpy_ndarray'}
 _INVERSE_ARRAY_TYPE_MAP = {'numpy_ndarray': _recreate_numpy_ndarray}
 if cuda_ndarray:

blocks/utils/__init__.py

@@ -10,13 +10,15 @@
 
 
 class UtilsModule(ModuleType):
-    utils_attrs = ("pack", "unpack", "reraise_as", "dict_subset", "dict_union",
-                   "repr_attrs", "ipdb_breakpoint", "print_sum", "print_shape",
-                   "change_recursion_limit", "extract_args", "find_bricks")
-    theano_utils_attrs = ("shared_floatx_zeros_matching", "shared_floatx_zeros",
-                          "shared_floatx_nans", "shared_floatx", "shared_like",
-                          "check_theano_variable", "is_graph_input",
-                          "is_shared_variable", "put_hook")
+    utils_attrs = (
+        "pack", "unpack", "reraise_as", "dict_subset", "dict_union",
+        "repr_attrs", "ipdb_breakpoint", "print_sum", "print_shape",
+        "change_recursion_limit", "extract_args", "find_bricks")
+    theano_utils_attrs = (
+        "shared_floatx_zeros_matching", "shared_floatx_zeros",
+        "shared_floatx_nans", "shared_floatx", "shared_like",
+        "check_theano_variable", "is_graph_input",
+        "is_shared_variable", "put_hook")
     __all__ = utils_attrs + theano_utils_attrs
     __doc__ = __doc__
     __file__ = __file__

blocks/utils/utils.py

@@ -4,7 +4,8 @@
 import contextlib
 from collections import OrderedDict, deque
 
-import numpy
+# for documentation
+import numpy  # noqa: F401
 
 
 def pack(arg):
@@ -362,4 +363,3 @@ def find_bricks(top_bricks, predicate):
                 found.append(current)
             to_visit.extend(current.children)
     return found
-

docs/conf.py

@@ -63,11 +63,13 @@
     'python': ('http://docs.python.org/3.4', None)
 }
 
+
 class Mock(MagicMock):
     @classmethod
     def __getattr__(cls, name):
             return Mock()
 
+
 MOCK_MODULES = ['fuel']
 sys.modules.update((mod_name, Mock()) for mod_name in MOCK_MODULES)
 

doctests/__init__.py

@@ -29,7 +29,7 @@ def load_tests(loader, tests, ignore):
             tests.addTests(doctest.DocTestSuite(
                 module=importlib.import_module(module), setUp=setup,
                 optionflags=doctest.IGNORE_EXCEPTION_DETAIL))
-        except:
+        except:  # noqa: E722
             pass
 
     # This part loads the doctests from the documentation

tests/algorithms/test_algorithms.py

@@ -186,8 +186,8 @@ def test_gradient_descent_spurious_sources():
     algorithm = GradientDescent(cost=cost, parameters=[W])
     algorithm.step_rule.learning_rate.set_value(0.75)
     algorithm.initialize()
-    assert_raises(lambda:
-                  algorithm.process_batch(dict(example_id='test')))
+    with assert_raises(ValueError):
+        algorithm.process_batch(dict(example_id='test'))
 
     algorithm = GradientDescent(cost=cost, parameters=[W],
                                 on_unused_sources='ignore')