Merge pull request #2798 from nouiz/tests

theano/gof/tests/test_vm.py

@@ -344,7 +344,9 @@ def test_reallocation():
     x = tensor.scalar('x')
     y = tensor.scalar('y')
     z = tensor.tanh(3 * x + y) + tensor.cosh(x + 5 * y)
-    for l in ['vm_nogc', 'vm', 'vm_nogc', 'vm']:
+    # The functinality is currently implement for non lazy and non c VM only.
+    for l in [vm.VM_Linker(allow_gc=False, lazy=False, use_cloop=False),
+              vm.VM_Linker(allow_gc=True, lazy=False, use_cloop=False)]:
         m = theano.compile.get_mode(theano.Mode(linker=l))
         m = m.excluding('fusion', 'inplace')
 

theano/tensor/tests/test_assert_op_gradient.py

-import numpy
-
-import theano
-import theano.tensor as T
-from theano.tensor.opt import Assert
-
-
-def test_assert_op_gradient():
-    x = T.vector('x')
-    assert_op = Assert()
-    cost = T.sum(assert_op(x, x.size < 2))
-    grad = T.grad(cost, x)
-    func = theano.function([x], grad)
-
-    x_val = numpy.ones(shape=(1,))
-    assert func(x_val) == 1

theano/tensor/tests/test_basic.py

@@ -235,7 +235,8 @@ def upcast_int8_nfunc(fn):
 
 def makeTester(name, op, expected, checks=None, good=None, bad_build=None,
                bad_runtime=None, grad=None, mode=None, grad_rtol=None,
-               eps=1e-10, skip=False, test_memmap=True, check_name=True):
+               eps=1e-10, skip=False, test_memmap=True, check_name=True,
+               grad_eps=None):
     """
     :param check_name:
         Use only for tester that aren't in Theano.
@@ -258,6 +259,7 @@ def makeTester(name, op, expected, checks=None, good=None, bad_build=None,
     _mode, _grad_rtol, _eps, skip_ = mode, grad_rtol, eps, skip
     _test_memmap = test_memmap
     _check_name = check_name
+    _grad_eps = grad_eps
 
     class Checker(unittest.TestCase):
 
@@ -463,8 +465,9 @@ def makeTester(name, op, expected, checks=None, good=None, bad_build=None,
                     inputs = [copy(input) for input in inputs]
                     try:
                         utt.verify_grad(self.op, inputs,
-                                mode=self.mode,
-                                rel_tol=_grad_rtol)
+                                        mode=self.mode,
+                                        rel_tol=_grad_rtol,
+                                        eps=_grad_eps)
                     except Exception, exc:
                         err_msg = ("Test %s::%s: Error occurred while"
                             " computing the gradient on the following"
@@ -976,6 +979,7 @@ ModTester = makeBroadcastTester(
     good=copymod(_good_broadcast_div_mod_normal_float,
                  ['complex1', 'complex2']),
     grad=_grad_broadcast_div_mod_normal,
+    grad_eps=1e-5,
     )
 
 
@@ -986,6 +990,7 @@ ModInplaceTester = makeBroadcastTester(
     good=copymod(_good_broadcast_div_mod_normal_float_inplace,
                  ["complex1", "complex2"]),
     grad=_grad_broadcast_div_mod_normal,
+    grad_eps=1e-5,
     inplace=True)
 
 _good_broadcast_pow_normal_float = dict(same_shapes=(rand_ranged(1, 5, (2, 3)), rand_ranged(-3, 3, (2, 3))),
@@ -2504,12 +2509,12 @@ class T_Clip(unittest.TestCase):
         rng = numpy.random.RandomState(utt.fetch_seed())
 
         nvals = 50
-        xval = rng.rand(nvals)
+        xval = rng.rand(nvals).astype(config.floatX)
         # To ensure that the min < x
-        yval_mn = rng.rand(nvals) - 1.0
+        yval_mn = rng.rand(nvals).astype(config.floatX) - 1.0
 
         # To ensure that the max > x
-        yval_mx = rng.rand(nvals) + 1.0
+        yval_mx = rng.rand(nvals).astype(config.floatX) + 1.0
 
         aval, = fn(xval, yval_mn)
         aval2, = fn2(xval, yval_mx)

theano/tensor/tests/test_extra_ops.py

@@ -369,12 +369,14 @@ class CompressTester(utt.InferShapeTester):
         self.op = compress
 
     def test_op(self):
-        for axis, cond, shape in zip(self.axis_list, self.cond_list, self.shape_list):
+        for axis, cond, shape in zip(self.axis_list, self.cond_list,
+                                     self.shape_list):
             cond_var = theano.tensor.ivector()
-            data     = numpy.random.random(size=shape).astype(theano.config.floatX)
+            data = numpy.random.random(size=shape).astype(theano.config.floatX)
             data_var = theano.tensor.matrix()
 
-            f = theano.function([cond_var, data_var], self.op(cond_var, data_var, axis=axis))
+            f = theano.function([cond_var, data_var],
+                                self.op(cond_var, data_var, axis=axis))
 
             expected = numpy.compress(cond, data, axis=axis)
             tested = f(cond, data)

theano/tensor/tests/test_opt.py

@@ -5053,7 +5053,7 @@ class Test_lift_transpose_through_dot(unittest.TestCase):
         a, b = matrices('ab')
         g = self.simple_optimize(FunctionGraph([a, b], [tensor.dot(a, b).T]))
         sg = '[dot(DimShuffle{1,0}(b), DimShuffle{1,0}(a))]'
-        assert str(g) == sg
+        assert str(g) == sg, (str(g), sg)
 
     def test_row_matrix(self):
         a = vector('a')
@@ -5063,7 +5063,7 @@ class Test_lift_transpose_through_dot(unittest.TestCase):
             [tensor.dot(a.dimshuffle('x', 0), b).T]),
             level='stabilize')
         sg = '[dot(DimShuffle{1,0}(b), DimShuffle{0,x}(a))]'
-        assert str(g) == sg
+        assert str(g) == sg, (str(g), sg)
 
     def test_matrix_col(self):
         a = vector('a')
@@ -5073,7 +5073,7 @@ class Test_lift_transpose_through_dot(unittest.TestCase):
             [tensor.dot(b, a.dimshuffle(0, 'x')).T]),
             level='stabilize')
         sg = '[dot(DimShuffle{x,0}(a), DimShuffle{1,0}(b))]'
-        assert str(g) == sg
+        assert str(g) == sg, (str(g), sg)
 
 
 def test_local_upcast_elemwise_constant_inputs():
@@ -5173,6 +5173,18 @@ class TestShapeFeature(unittest.TestCase):
         self.assertRaises(IndexError, shape_feature.same_shape, x, o, 1, 0)
         self.assertRaises(IndexError, shape_feature.same_shape, x, o, 0, 1)
 
+
+def test_assert_op_gradient():
+    x = T.vector('x')
+    assert_op = Assert()
+    cost = T.sum(assert_op(x, x.size < 2))
+    grad = T.grad(cost, x)
+    func = theano.function([x], grad)
+
+    x_val = numpy.ones(shape=(1,), dtype=theano.config.floatX)
+    assert func(x_val) == 1
+
+
 if __name__ == '__main__':
     t = TestMakeVector('setUp')
     t.setUp()