Merge pull request #1348 from nouiz/fixes

theano/sandbox/linalg/tests/test_linalg.py

@@ -536,6 +536,8 @@ class test_Eigh_float32(test_Eigh):
     dtype = 'float32'
 
 def test_matrix_inverse_solve():
+    if not imported_scipy:
+        raise SkipTest("Scipy needed for the Solve op.")
     A = theano.tensor.dmatrix('A')
     b = theano.tensor.dmatrix('b')
     node = matrix_inverse(A).dot(b).owner

theano/tensor/basic.py

@@ -366,9 +366,14 @@ def constant_or_value(x, rtype, name=None, ndim=None, dtype=None):
             # Theano graph, because on Windows 64, all shapes are expressed
             # with longs.
             # If a long fits in int64, we convert it into an int64, like
-            # numpy.asarray() does.
+            # numpy.asarray() does up to 1.7. NumPy 1.7.1 upcaset to int64
+            # if possible, but fallback to uint64 if int64 isn't possible but
+            # uint64 is. We always do as NumPy 1.7.1  here.
             # If x is too big, an OverflowError will be raised by numpy.
+            try:
                 x_ = theano._asarray(x, dtype='int64')
+            except OverflowError:
+                x_ = theano._asarray(x, dtype='uint64')
         elif isinstance(x, numpy.ndarray):
             x_ = x
             # Currently we do not have a bool dtype in Theano.
@@ -377,6 +382,10 @@ def constant_or_value(x, rtype, name=None, ndim=None, dtype=None):
             if x.dtype == 'bool':
                 x_ = numpy.asarray(x_, dtype='uint8')
         else:
+            # Here x is probably a list or a tuple. If it contain a long,
+            # we will behave like the current NumPy version: 1.7 and bellow,
+            # it will only work if the long fit in int64. For NumPy 1.7.1+,
+            # it will work if the long git in int64 or uint64.
             x_ = numpy.asarray(x)
 
     assert type(x_) == numpy.ndarray

theano/tensor/tests/test_basic.py

@@ -2016,7 +2016,6 @@ def test_identity():
 
 class CastTester(unittest.TestCase):
     def test_good_between_real_types(self):
-        expected = lambda x, y: x.astype(y),
         good = itertools.chain(
                     multi_dtype_cast_checks((2,), dtypes=REAL_DTYPES),
                     # Casts from foo to foo
@@ -2029,6 +2028,10 @@ class CastTester(unittest.TestCase):
             f = function([inp], out)
             assert f(obj).dtype == numpy.dtype(dtype)
 
+            # Test astype too
+            out2 = inp.astype(dtype=dtype)
+            assert out2.type == out.type
+
     def test_cast_from_real_to_complex(self):
         for real_dtype in REAL_DTYPES:
             for complex_dtype in COMPLEX_DTYPES:
@@ -2313,7 +2316,6 @@ class T_max_and_argmax(unittest.TestCase):
 
         x = matrix()
         cost = argmax(x, axis=0).sum()
-        value_error_raised = False
         gx = grad(cost, x)
         val = tensor.get_scalar_constant_value(gx)
         assert val == 0.0
@@ -2916,11 +2918,8 @@ class T_subtensor(unittest.TestCase, utt.TestOptimizationMixin):
                 oldlevel = _logger.level
                 _logger.setLevel(logging.CRITICAL)
                 try:
-                    try:
-                        tval = self.eval_output_and_check([t])
-                        assert 0
-                    except IndexError, e:
-                        pass
+                    self.assertRaises(IndexError,
+                                      self.eval_output_and_check, [t])
                 finally:
                     _logger.setLevel(oldlevel)
         finally:
@@ -2928,16 +2927,13 @@ class T_subtensor(unittest.TestCase, utt.TestOptimizationMixin):
 
     def test2_err_bounds1(self):
         n = self.shared((numpy.ones((2, 3), dtype=self.dtype) * 5))
-        t = n[4:5, 2]
+        t = n[4:5, 3]
         self.assertTrue(isinstance(t.owner.op, Subtensor))
         old_stderr = sys.stderr
         sys.stderr = StringIO.StringIO()
         try:
-            try:
-                tval = self.eval_output_and_check([t])
-            except Exception, e:
-                if e[0] != 'index out of bounds':
-                    raise
+            self.assertRaises(IndexError,
+                              self.eval_output_and_check, [t])
         finally:
             sys.stderr = old_stderr
 
@@ -4313,7 +4309,7 @@ class T_Join_and_Split(unittest.TestCase):
         assert (out == [3, 13]).all()
 
         if theano.config.mode != 'FAST_COMPILE':
-            for node in f.maker.fgraph.toposort():
+            for node in topo:
                 assert not isinstance(node.op, tensor.Join)
 
         # Test hide error
@@ -5161,7 +5157,6 @@ class T_reshape(unittest.TestCase):
     def test_bad_shape(self):
         a = matrix('a')
         shapes = ivector('shapes')
-        ndim = 2
         rng = numpy.random.RandomState(seed=utt.fetch_seed())
         a_val = rng.uniform(size=(3, 4)).astype(config.floatX)
 
@@ -5172,7 +5167,7 @@ class T_reshape(unittest.TestCase):
         f = function([a, shapes], z.shape)
         self.assertRaises(ValueError, f, a_val, [13])
 
-        #Test reshape to 1 dim
+        #Test reshape to 2 dim
         r = a.reshape(shapes, ndim=2)
         z = zeros_like(r)
 
@@ -6392,6 +6387,30 @@ class T_long_tensor(unittest.TestCase):
 
     def test_too_big(self):
         val = 2L ** 63
+        #NumPy 1.7 this will raise an exception
+        #NumPy 1.7.1 this will work
+        try:
+            cst = constant(val)
+            assert cst.value == val
+            assert cst.dtype == "uint64"
+        except OverflowError:
+            pass
+
+        try:
+            cst = constant([val, val])
+            assert cst.value == val
+            assert cst.dtype == "uint64"
+        except TypeError:
+            pass
+        try:
+            cst = constant([[val, val]])
+            assert cst.value == val
+            assert cst.dtype == "uint64"
+        except TypeError:
+            pass
+
+        val = 2L ** 64
+        # This fail for all NumPy version.
         self.assertRaises(Exception, constant, val)
         self.assertRaises(Exception, constant, [val, val])
         self.assertRaises(Exception, constant, [[val, val]])
@@ -7055,7 +7074,6 @@ class TestInferShape(utt.InferShapeTester):
 
         # ScalarFromTensor
         aiscal = iscalar()
-        aconst = constant(45)
         self._compile_and_check([aiscal],
                             [TensorFromScalar()(ScalarFromTensor()(aiscal))],
                                 [45], ScalarFromTensor,