Merge pull request #3368 from kashif/issue-3347

theano/sandbox/cuda/dnn.py

@@ -2321,10 +2321,10 @@ if True:
                 (inp_grad.owner and isinstance(inp_grad.owner.op,
                                                HostFromGpu))):
                 desc = GpuDnnPoolDesc(ws=ds, stride=st, mode=mode, pad=pad)()
+                contiguous_inp_grad = gpu_contiguous(inp_grad)
                 ret = GpuDnnPoolGrad()(gpu_contiguous(inp),
-                                       gpu_contiguous(numpy.empty((1,1,1,1),
-                                           dtype=numpy.float32)),
-                                       gpu_contiguous(inp_grad),
+                                       contiguous_inp_grad,
+                                       contiguous_inp_grad,
                                        desc)
                 return [host_from_gpu(ret)]
 

theano/sandbox/cuda/tests/test_dnn.py

@@ -450,24 +450,29 @@ def test_pooling_opt():
     if not cuda.dnn.dnn_available():
         raise SkipTest(cuda.dnn.dnn_available.msg)
 
-    x = T.ftensor4()
+    x = T.fmatrix()
 
     f = theano.function(
         [x],
-        max_pool_2d(x, ds=(2, 2), ignore_border=True),
+        max_pool_2d(x, ds=(2, 2), mode='average_inc_pad', ignore_border=True),
         mode=mode_with_gpu)
 
     assert any([isinstance(n.op, cuda.dnn.GpuDnnPool)
                 for n in f.maker.fgraph.toposort()])
 
+    f(numpy.zeros((10, 10), dtype='float32'))
+
     f = theano.function(
         [x],
-        T.grad(max_pool_2d(x, ds=(2, 2), ignore_border=True).sum(), x),
+        T.grad(max_pool_2d(x, ds=(2, 2), mode='average_inc_pad',
+                           ignore_border=True).sum(), x),
         mode=mode_with_gpu.including("cudnn"))
 
     assert any([isinstance(n.op, cuda.dnn.GpuDnnPoolGrad)
                 for n in f.maker.fgraph.toposort()])
 
+    f(numpy.zeros((10, 10), dtype='float32'))
+
 
 class test_DnnSoftMax(test_nnet.test_SoftMax):
     gpu_op = dnn.GpuDnnSoftmax

theano/sandbox/gpuarray/dnn.py

@@ -1763,9 +1763,10 @@ def local_avg_pool_dnn_grad_stride(node):
     mode = node.op.mode
 
     desc = GpuDnnPoolDesc(ws=ds, stride=st, mode=mode, pad=pad)()
+    contiguous_inp_grad = gpu_contiguous(inp_grad)
     return GpuDnnPoolGrad()(gpu_contiguous(inp),
-                            gpu_contiguous(numpy.empty((1, 1, 1, 1), dtype=numpy.float32)),
-                            gpu_contiguous(inp_grad),
+                            contiguous_inp_grad,
+                            contiguous_inp_grad,
                             desc)
 
 

theano/sandbox/gpuarray/tests/test_dnn.py

@@ -278,24 +278,31 @@ def test_pooling_opt():
     if not dnn.dnn_available():
         raise SkipTest(dnn.dnn_available.msg)
 
-    x = T.ftensor4()
+    x = T.fmatrix()
 
     f = theano.function(
         [x],
-        max_pool_2d(x, ds=(2, 2), ignore_border=True),
+        max_pool_2d(x, ds=(2, 2), mode='average_inc_pad',
+                    ignore_border=True),
         mode=mode_with_gpu)
 
     assert any([isinstance(n.op, dnn.GpuDnnPool)
                 for n in f.maker.fgraph.toposort()])
 
+    f(numpy.zeros((10, 10), dtype='float32'))
+
     f = theano.function(
         [x],
-        T.grad(max_pool_2d(x, ds=(2, 2), ignore_border=True).sum(), x),
+        T.grad(max_pool_2d(x, ds=(2, 2), mode='average_inc_pad',
+                           ignore_border=True).sum(),
+               x),
         mode=mode_with_gpu.including("cudnn"))
 
     assert any([isinstance(n.op, dnn.GpuDnnPoolGrad)
                 for n in f.maker.fgraph.toposort()])
 
+    f(numpy.zeros((10, 10), dtype='float32'))
+
 
 def test_dnn_tag():
     """