Merge pull request #801 from lamblin/fix_careduce_nan

theano/scalar/basic.py

@@ -753,7 +753,7 @@ class ScalarOp(Op):
                 return self.__class__.__name__
 
     def c_code_cache_version(self):
-        return (3,)
+        return (4,)
 
 
 class UnaryScalarOp(ScalarOp):
@@ -1078,7 +1078,9 @@ class Maximum(BinaryScalarOp):
     def c_code(self, node, name, (x, y), (z, ), sub):
         if any([i.type in complex_types for i in node.inputs]):
             raise NotImplementedError()
-        return "%(z)s = ((%(y)s)>(%(x)s)? (%(y)s):(%(x)s));" % locals()
+        # Test for both y>x and x>=y to detect NaN
+        return ('%(z)s = ((%(y)s)>(%(x)s)? (%(y)s): '
+                    '((%(x)s)>=(%(y)s)? (%(x)s): nan("")));' % locals())
 
     def grad(self, (x, y), (gz, )):
         assert gz.type not in complex_types
@@ -1103,7 +1105,8 @@ class Minimum(BinaryScalarOp):
     def c_code(self, node, name, (x, y), (z, ), sub):
         if any([i.type in complex_types for i in node.inputs]):
             raise NotImplementedError()
-        return "%(z)s = ((%(y)s)<(%(x)s)? (%(y)s):(%(x)s));" % locals()
+        return ('%(z)s = ((%(y)s)<(%(x)s)? (%(y)s): '
+                    '((%(x)s)<=(%(y)s)? (%(x)s): nan("")));' % locals())
 
     def grad(self, (x, y), (gz, )):
         assert gz.type not in complex_types

theano/tensor/elemwise.py

@@ -1209,8 +1209,12 @@ class CAReduce(Op):
                 # if available
                 if variable.shape[dimension] == 0:
                     if hasattr(self.scalar_op, 'identity'):
-                        variable = numpy.array(self.scalar_op.identity)
-                        break
+                        # Compute the shape of the output
+                        v_shape = list(variable.shape)
+                        del v_shape[dimension]
+                        variable = numpy.empty(tuple(v_shape),
+                                               dtype=variable.dtype)
+                        variable.fill(self.scalar_op.identity)
                     else:
                         raise ValueError((
                             "Input (%s) has zero-size on axis %s, but "

theano/tensor/tests/test_elemwise.py

@@ -315,15 +315,18 @@ class test_CAReduce(unittest_tools.InferShapeTester):
                 else:
                     self.fail()
             else:
-                #numpy.{all,any} return bool type.
+                # numpy.{all,any} return bool type,
+                # but theano ops return an int8 array instead
                 if scalar_op in [scalar.and_, scalar.or_]:
-                    zv = numpy.asarray(zv, dtype=dtype)
+                    zv = numpy.asarray(zv, dtype='int8')
                 if test_nan:
                     self.assertTrue(theano.tensor.TensorType.values_eq(f(xv),
                                                                        zv),
                                     (f(xv), zv))
                 else:
-                    self.assertTrue(numpy.allclose(f(xv), zv), (f(xv), zv))
+                    f_xv = f(xv)
+                    self.assertTrue((f_xv.shape == zv.shape), (f_xv, zv))
+                    self.assertTrue(numpy.allclose(f_xv, zv), (f_xv, zv))
 
             #test CAReduce.infer_shape
             #the Shape op don't implement c_code!
@@ -355,10 +358,6 @@ class test_CAReduce(unittest_tools.InferShapeTester):
             self.with_linker(gof.PerformLinker(), scalar.and_, dtype=dtype)
             self.with_linker(gof.PerformLinker(), scalar.xor, dtype=dtype)
 
-    @dec.knownfailureif(
-        True,
-        ("When there is nan in the input of CAReduce,"
-         " we don't have a good output. "))
     def test_perform_nan(self):
         for dtype in ["floatX", "complex64", "complex128"]:
             self.with_linker(gof.PerformLinker(), scalar.add, dtype=dtype,
@@ -370,12 +369,8 @@ class test_CAReduce(unittest_tools.InferShapeTester):
             self.with_linker(gof.PerformLinker(), scalar.minimum, dtype=dtype,
                              test_nan=True)
             self.with_linker(gof.PerformLinker(), scalar.or_, dtype=dtype,
-                             test_nan=True)
-            self.with_linker(gof.PerformLinker(), scalar.and_, dtype=dtype,
-                             test_nan=True)
-            self.with_linker(gof.PerformLinker(), or_, dtype=dtype,
                              test_nan=True, tensor_op=tensor.any)
-            self.with_linker(gof.PerformLinker(), and_, dtype=dtype,
+            self.with_linker(gof.PerformLinker(), scalar.and_, dtype=dtype,
                              test_nan=True, tensor_op=tensor.all)
 
     def test_c(self):
@@ -394,10 +389,6 @@ class test_CAReduce(unittest_tools.InferShapeTester):
             self.with_linker(gof.CLinker(), scalar.and_, dtype=dtype)
             self.with_linker(gof.CLinker(), scalar.xor, dtype=dtype)
 
-    @dec.knownfailureif(
-        True,
-        ("When there is nan in the input of CAReduce,"
-         " we don't have a good output. "))
     def test_c_nan(self):
         for dtype in ["floatX", "complex64", "complex128"]:
             self.with_linker(gof.CLinker(), scalar.add, dtype=dtype,