fix the gpu conv optimizer when the cpu and gpu conv op don't have the same pattern.

theano/sandbox/cuda/opt.py

@@ -497,7 +497,9 @@ def local_gpu_conv(node):
         if host_input.owner and isinstance(host_input.owner.op, conv.ConvOp):
             gpu_conv = GpuConvOp_from_ConvOp(host_input.owner.op)
             img, kern = host_input.owner.inputs
-            return [gpu_conv(gpu_from_host(img), gpu_from_host(kern))]
+            #in some case the ConvOp broadcast the last 2 dimensions differently then the gpu ConvOp
+            return [tensor.patternbroadcast(gpu_conv(gpu_from_host(img), gpu_from_host(kern)),
+                                            node.outputs[0].broadcastable)]
 
     if isinstance(node.op, conv.ConvOp):
         #conv(host_from_gpu) -> host_from_gpu(gpu_conv)
@@ -506,7 +508,10 @@ def local_gpu_conv(node):
         kern_on_gpu = (kern.owner and kern.owner.op == host_from_gpu)
         if img_on_gpu or kern_on_gpu:
             gpu_conv = GpuConvOp_from_ConvOp(node.op)
-            return [host_from_gpu(gpu_conv(gpu_from_host(img), gpu_from_host(kern)))]
+            #in some case the ConvOp broadcast the last 2 dimensions differently then the gpu ConvOp
+            return [tensor.patternbroadcast(host_from_gpu(gpu_conv(gpu_from_host(img),
+                                                                   gpu_from_host(kern))),
+                                            node.outputs[0].broadcastable)]
 
 import theano.tensor.signal.downsample as downsample
 @register_opt()

theano/tensor/basic.py

@@ -2744,6 +2744,14 @@ def unbroadcast(x, *axes):
     rval = Rebroadcast(*[(axis, False) for axis in axes])(x)
     return theano.tensor.opt.apply_rebroadcast_opt(rval)
 
+def patternbroadcast(x, broadcastable):
+    """
+    Make the input impossible to broadcast in the specified axes.
+    
+    We apply the opt here to don't pollute the graph especially during the gpu optimization
+    """
+    rval = Rebroadcast(*[(i,broadcastable[i]) for i in range(len(broadcastable))])(x)
+    return theano.tensor.opt.apply_rebroadcast_opt(rval)
 
 class Join(Op):
     """