Only run dnnconv tests with dilation if v>=6.

theano/gpuarray/dnn.py

@@ -384,6 +384,9 @@ class GpuDnnConvDesc(COp):
                  precision="float32"):
         COp.__init__(self, ["conv_desc.c"], "APPLY_SPECIFIC(conv_desc)")
 
+        if version() < 6000 and any([d != 1 for d in dilation]):
+            raise RuntimeError("Dilation > 1 not supported for cuDNN version < 6.")
+
         if isinstance(border_mode, integer_types):
             border_mode = (border_mode,) * len(subsample)
         if isinstance(border_mode, tuple):
@@ -455,17 +458,12 @@ class GpuDnnConvDesc(COp):
         else:
             sub2 = '0'
 
-        if version() < 6000:
-            dil0 = '1'
-            dil1 = '1'
-            dil2 = '1'
+        dil0 = str(self.dilation[0])
+        dil1 = str(self.dilation[1])
+        if len(self.dilation) > 2:
+            dil2 = str(self.dilation[2])
         else:
-            dil0 = str(self.dilation[0])
-            dil1 = str(self.dilation[1])
-            if len(self.dilation) > 2:
-                dil2 = str(self.dilation[2])
-            else:
-                dil2 = '0'
+            dil2 = '0'
 
         if self.precision == 'float16':
             precision = 'CUDNN_DATA_HALF'

theano/gpuarray/tests/test_dnn.py

@@ -641,7 +641,7 @@ class TestDnnInferShapes(utt.InferShapeTester):
     def test_conv(self, algo, border_mode, conv_mode):
         # Currently only CUDNN_CONVOLUTION_FWD_ALGO_IMPLICIT_GEMM (algo 'none')
         # supports dilation > 1.
-        dilations = [(1, 1), (2, 2)] if algo == "none" else [(1, 1)]
+        dilations = [(1, 1), (2, 2)] if (algo == "none" and dnn.version() >= 6000) else [(1, 1)]
 
         self._test_conv(T.tensor4('img'),
                         T.tensor4('kerns'),
@@ -712,6 +712,7 @@ class TestDnnInferShapes(utt.InferShapeTester):
 
     @parameterized.expand(product(border_modes, conv_modes), utt.custom_name_func)
     def test_conv_gradw(self, border_mode, conv_mode):
+        dilations = [(1, 1), (2, 2)] if dnn.version() >= 6000 else [(1, 1)]
         self._test_conv_gradw(T.tensor4('img'),
                               T.tensor4('topgrad'),
                               T.tensor4('kerns'),
@@ -720,7 +721,7 @@ class TestDnnInferShapes(utt.InferShapeTester):
                               border_mode,
                               conv_mode,
                               [(1, 1)],
-                              [(1, 1), (2, 2)])
+                              dilations)
 
     def test_conv_gradi(self):
         if not dnn.dnn_available(test_ctx_name):
@@ -737,10 +738,11 @@ class TestDnnInferShapes(utt.InferShapeTester):
             dtype=theano.config.floatX
         )
 
+        dilations = [(1, 1), (2, 2)] if dnn.version() >= 6000 else [(1, 1)]
         for border_mode, subsample, dilation, conv_mode in product(
             ['valid', 'full'],
             [(1, 1)],
-            [(1, 1), (2, 2)],
+            dilations,
             ['conv', 'cross']
         ):
             shape = get_conv_gradinputs_shape(kern_vals.shape, out_vals.shape, border_mode, subsample, dilation)