small fix for tests

theano/tensor/nnet/tests/test_corr.py

@@ -430,14 +430,14 @@ class TestGroupCorr2d(Grouped_conv_noOptim):
     def test_graph(self):
         # define common values  first
         groups = 3
-        bottom = np.random.rand(3, 6, 5, 5)
-        kern = np.random.rand(9, 2, 3, 3)
+        bottom = np.random.rand(3, 6, 5, 5).astype('float32')
+        kern = np.random.rand(9, 2, 3, 3).astype('float32')
         bottom_sym = T.tensor4('bottom')
         kern_sym = T.tensor4('kern')
 
         # grouped convolution graph
         conv_group = self.conv2d(num_groups=groups)(bottom_sym, kern_sym)
-        gconv_func = theano.function([bottom_sym, kern_sym], conv_group)
+        gconv_func = theano.function([bottom_sym, kern_sym], conv_group, mode=self.mode)
 
         # Graph for the normal hard way
         kern_offset = kern_sym.shape[0] // groups
@@ -446,7 +446,7 @@ class TestGroupCorr2d(Grouped_conv_noOptim):
                              kern_sym[i * kern_offset:(i + 1) * kern_offset, :, :, :])
                              for i in range(groups)]
         concatenated_output = T.concatenate(split_conv_output, axis=1)
-        conv_func = theano.function([bottom_sym, kern_sym], concatenated_output)
+        conv_func = theano.function([bottom_sym, kern_sym], concatenated_output, mode=self.mode)
 
         # calculate outputs for each graph
         gconv_output = gconv_func(bottom, kern)