move TypeError into make_node, consequently changed the test a little. Add a…

theano/tensor/extra_ops.py

@@ -760,6 +760,11 @@ class FillDiagonalOffset(gof.Op):
         if val.dtype != a.dtype:
             raise TypeError('%s: type of second parameter must be the same'
                             ' as the first\'s' % self.__class__.__name__)
+        elif offset.dtype[:3] != 'int':
+            raise TypeError('%s: type of third parameter must be as integer'
+                            ' use theano.tensor.cast( input, \'int32/int64\')' \
+                            % self.__class__.__name__)
+
 
 
         return gof.Apply(self, [a, val, offset], [a.type()])
@@ -770,11 +775,6 @@ class FillDiagonalOffset(gof.Op):
         offset = inputs[2]
         height, width = a.shape
 
-        # offset should be an integer
-        if offset % 1 != 0:
-            raise ValueError('%s: third parameter must be an integer'\
-                            % self.__class__.__name__)
-
         """
         Note: The fill_diagonal only support rectangular matrix. The output
         of tall matrix is "wrapped", which is an option in numpy 1.9.0

theano/tensor/tests/test_extra_ops.py

@@ -422,11 +422,11 @@ class TestFillDiagonalOffset(utt.InferShapeTester):
         x = tensor.matrix()
         y = tensor.scalar()
         z = tensor.scalar()
+        z_in = tensor.cast( z, "int32")
 
-        test_offset = numpy.array(numpy.random.randint(-5,5),
-            dtype = config.floatX)
+        test_offset = numpy.random.randint(-5,5)
 
-        f = function([x, y, z], fill_diagonal_offset(x, y, z))
+        f = function([x, y, z_in], fill_diagonal_offset(x, y, z_in))
         for shp in [(8, 8), (5, 8), (8, 5)]:
             a = numpy.random.rand(*shp).astype(config.floatX)
             val = numpy.cast[config.floatX](numpy.random.rand())
@@ -443,8 +443,7 @@ class TestFillDiagonalOffset(utt.InferShapeTester):
                                         a.shape[0]+test_offset )
 
     def test_gradient(self):
-        test_offset = numpy.array(numpy.random.randint(-5,5),
-                        dtype = config.floatX)
+        test_offset = numpy.random.randint(-5,5)
         # input 'offset' will not be tested
         def fill_diagonal_with_fix_offset( a, val):
             return fill_diagonal_offset( a, val, test_offset)
@@ -460,10 +459,9 @@ class TestFillDiagonalOffset(utt.InferShapeTester):
         x = tensor.dmatrix()
         y = tensor.dscalar()
         z = tensor.dscalar()
-        test_offset = numpy.array(numpy.random.randint(-5,5),
-                        dtype = config.floatX)
-        self._compile_and_check([x, y, z], [self.op(x, y, z)],
+        z_in = tensor.cast( z, "int32")
+        self._compile_and_check([x, y, z_in], [self.op(x, y, z_in)],
                                 [numpy.random.rand(8, 5),
                                  numpy.random.rand(),
-                                 test_offset],
-                                self.op_class)
+                                 numpy.random.randint(-5,5)],
+                                 self.op_class )