Merge pull request #4033 from nouiz/py3_repr

theano/compile/debugmode.py

@@ -2050,6 +2050,7 @@ class _Linker(gof.link.LocalLinker):
                             # shouldn't have put it into the list in
                             # the first place
                             thunk_py = None
+                            thunks_py[i] = None
                         except Exception as e:
                             # I think that only 1 optimization can
                             # insert a given apply node. If that is not True,

theano/sandbox/cuda/blas.py

@@ -1143,6 +1143,8 @@ class GpuCorrMM_gradWeights(BaseGpuCorrMM):
                 raise ValueError('shape must be given if subsample != (1, 1)'
                                  ' or border_mode == "half"')
             height_width = [shape[0], shape[1]]
+            assert shape[0].ndim == 0
+            assert shape[1].ndim == 0
         else:
             height_width = []
 
@@ -1201,6 +1203,9 @@ class GpuCorrMM_gradInputs(BaseGpuCorrMM):
         if self.subsample != (1, 1) and shape is None:
             raise ValueError('shape must be given if subsample != (1, 1)')
         height_width = [shape[0], shape[1]] if self.subsample != (1, 1) else []
+        if height_width:
+            assert shape[0].ndim == 0
+            assert shape[1].ndim == 0
 
         broadcastable = [topgrad.type.broadcastable[0], kern.type.broadcastable[1],
                          False, False]

theano/sandbox/cuda/cuda_ndarray.cu

@@ -3066,6 +3066,12 @@ PyObject *CudaNdarray_repr(PyObject *self)
                         ")");
     Py_DECREF(str);
     Py_DECREF(np_object);
+    #if PY_MAJOR_VERSION >= 3
+    // In Python 3 PyString_FromFormat return a Bytes object
+    PyObject* out2 = PyObject_Str(out);
+    Py_DECREF(out);
+    return out2;
+    #endif
     return out;
 }
 

theano/sandbox/cuda/opt.py

@@ -2767,10 +2767,11 @@ def local_abstractconv_gemm(node):
         # GpuConv does not always store information on the batchsize and
         # channels, though, so we only use what information we have.)
         if ((subsample == (1, 1)) and
-                (node.op.imshp is not None) and
-                (None not in node.op.imshp[-2:]) and
-                (node.op.kshp is not None) and
-                (None not in node.op.kshp)):
+            (node.op.imshp is not None) and
+            (None not in node.op.imshp[-2:]) and
+            (node.op.kshp is not None) and
+            (None not in node.op.kshp) and
+             border_mode != "half"):
             # we know the kernel and output size
             prod1 = node.op.kshp[0] * node.op.kshp[1]
             prod2 = ((node.op.imshp[-2] - node.op.kshp[0] + 1) *

theano/sandbox/cuda/tests/test_opt.py

@@ -859,6 +859,20 @@ class Test_GpuReshape(test_opt.Test_Reshape):
         self.op = basic_ops.GpuReshape
 
 
+def test_local_abstractconv_gemm():
+    """ We test it here as this is the optimization only that we test.
+    This test gh-4036"""
+    image = tensor.ftensor4()
+    W = tensor.ftensor4()
+    conv = tensor.nnet.conv2d(image,
+                         W,
+                         input_shape=(1, 32, 32, 32),
+                         filter_shape=(32, 32, 3, 3),
+                         border_mode='half')
+    f = theano.function([image, W], [conv], mode=mode_with_gpu)
+    f(numpy.random.rand(1, 32, 32, 32).astype('float32'),
+      numpy.random.rand(32, 32, 3, 3).astype('float32'))
+
 if __name__ == '__main__':
     test_gpualloc()
     test_opt_gpujoin_onlyajoin()