Fixed some tests when config.cast_policy == numpy+floatX and config.floatX == float32

theano/tensor/tests/test_basic.py

@@ -201,7 +201,7 @@ def makeTester(name, op, expected, checks = {}, good = {}, bad_build = {},
                 if not isinstance(expecteds, (list, tuple)):
                     expecteds = (expecteds, )
 
-                for i, (variable, expected) in enumerate(zip(variables, expecteds)):
+                for i, (variable, expected) in enumerate(izip(variables, expecteds)):
                     if variable.dtype != expected.dtype or variable.shape != expected.shape or \
                             numpy.any(numpy.abs(variable - expected) > eps):
                         self.fail("Test %s::%s: Output %s gave the wrong value. With inputs %s, expected %s, got %s. numpy.allclose return %s %s"
@@ -347,8 +347,31 @@ _grad_broadcast_binary_normal = dict(same_shapes = (rand(2, 3), rand(2, 3)),
                                      )
 
 
+def check_floatX(inputs, rval):
+    """
+    :param inputs: Inputs to a function that returned `rval` with these inputs.
+
+    :param rval: Value returned by a function with inputs set to `inputs`.
+
+    :returns: Either `rval` unchanged, or `rval` cast in float32. The idea is
+    that when a numpy function would have returned a float64, Theano may prefer
+    to return a float32 instead when `config.cast_policy` is set to
+    'numpy+floatX' and config.floatX to 'float32', and there was no float64
+    input.
+    """
+    if (isinstance(rval, numpy.ndarray) and
+        rval.dtype == 'float64' and
+        config.cast_policy == 'numpy+floatX'
+        and config.floatX == 'float32' and
+        all(x.dtype != 'float64' for x in inputs)):
+        # Then we expect float32 instead of float64.
+        return rval.astype('float32')
+    else:
+        return rval
+
+
 AddTester = makeBroadcastTester(op = add,
-                                  expected = lambda *inputs: reduce(lambda x, y: x + y, inputs),
+                                  expected = lambda *inputs: check_floatX(inputs, reduce(lambda x, y: x + y, inputs)),
                                   good = dict(three_inputs_same_shapes = (rand(2, 3), rand(2, 3), rand(2, 3)),
                                               four_inputs_broadcast = (rand(2, 3), rand(1, 3), rand(2, 1), rand(1, 1)),
                                               **_good_broadcast_binary_normal),
@@ -364,7 +387,7 @@ AddInplaceTester = makeBroadcastTester(op = inplace.add_inplace,
                                          inplace = True)
 
 SubTester = makeBroadcastTester(op = sub,
-                                  expected = lambda x, y: x - y,
+                                  expected = lambda x, y: check_floatX((x, y), x - y),
                                   good = _good_broadcast_binary_normal,
                                   bad_build = _bad_build_broadcast_binary_normal,
                                   bad_runtime = _bad_runtime_broadcast_binary_normal,
@@ -379,7 +402,7 @@ SubInplaceTester = makeBroadcastTester(op = inplace.sub_inplace,
                                          inplace = True)
 
 MaximumTester = makeBroadcastTester(op = maximum,
-                                  expected = numpy.maximum,
+                                  expected = lambda *inputs: check_floatX(inputs, numpy.maximum(*inputs)),
                                   good = _good_broadcast_binary_normal,
                                   bad_build = _bad_build_broadcast_binary_normal,
                                   bad_runtime = _bad_runtime_broadcast_binary_normal,
@@ -394,7 +417,7 @@ MaximumInplaceTester = makeBroadcastTester(op = inplace.maximum_inplace,
                                          inplace = True)
 
 MinimumTester = makeBroadcastTester(op = minimum,
-                                  expected = numpy.minimum,
+                                  expected = lambda *inputs: check_floatX(inputs, numpy.minimum(*inputs)),
                                   good = _good_broadcast_binary_normal,
                                   bad_build = _bad_build_broadcast_binary_normal,
                                   bad_runtime = _bad_runtime_broadcast_binary_normal,
@@ -409,7 +432,7 @@ MinimumInplaceTester = makeBroadcastTester(op = inplace.minimum_inplace,
                                          inplace = True)
 
 MulTester = makeBroadcastTester(op = mul,
-                                  expected = lambda *inputs: reduce(lambda x, y: x * y, inputs),
+                                  expected = lambda *inputs: check_floatX(inputs, reduce(lambda x, y: x * y, inputs)),
                                   good = dict(three_inputs_same_shapes = (rand(2, 3), rand(2, 3), rand(2, 3)),
                                               four_inputs_broadcast = (rand(2, 3), rand(1, 3), rand(2, 1), rand(1, 1)),
                                               **_good_broadcast_binary_normal),
@@ -475,7 +498,7 @@ if config.floatX=='float32':
     # This is probably caused by our way of computing the gradient error.
     div_grad_rtol=0.025
 DivTester = makeBroadcastTester(op = true_div,
-                                  expected = lambda x, y: x / y,
+                                  expected = lambda x, y: check_floatX((x, y), x / y),
                                   good = _good_broadcast_div_mod_normal_float,
 #                                               integers = (randint(2, 3), randint_nonzero(2, 3)),
 #                                               dtype_mixup_1 = (rand(2, 3), randint_nonzero(2, 3)),
@@ -527,7 +550,7 @@ _good_broadcast_pow_normal_float_pow = copy(_good_broadcast_pow_normal_float)
 del _good_broadcast_pow_normal_float_pow["empty2"]
 
 PowTester = makeBroadcastTester(op = pow,
-                                  expected = lambda x, y: x ** y,
+                                  expected = lambda x, y: check_floatX((x, y), x ** y),
                                   good = _good_broadcast_pow_normal_float,
                                   grad = _grad_broadcast_pow_normal)
 PowInplaceTester = makeBroadcastTester(op = inplace.pow_inplace,