Fix some things and remove extraneous dimshuffles.

theano/gpuarray/dnn.py

@@ -1982,7 +1982,7 @@ class _RNNSplitParams(DnnBase):
         assert w.ndim == 1
         layer = as_scalar(layer).astype('int32')
         isize = as_tensor_variable(isize).astype('uint64')
-        assert isize.ndim == 2
+        assert isize.ndim == 1
         typecode = as_scalar(typecode).astype('int32')
         _1d = GpuArrayType(w.type.dtype, [False],
                            context_name=w.type.context_name)
@@ -2230,7 +2230,7 @@ class GpuDnnRNNOp(DnnBase):
 
         return Apply(self, inputs, outputs)
 
-    def grad2(self, inputs, outputs, output_grads):
+    def L_op(self, inputs, outputs, output_grads):
         desc, w, x, hx = inputs[:4]
         cx = inputs[4] if len(inputs) == 5 else None
         reserve, y, hy = outputs[:3]

theano/gpuarray/tests/rnn_support.py

@@ -73,14 +73,14 @@ class Layer(object):
 
 
 class GRU(Layer):
-    def __init__(self, input_dim, output_dim, input_layer, s0=None, batch_normalize=False, name=""):
+    def __init__(self, input_dim, output_dim, input_layer, s0=None, name=""):
         '''Layers information'''
         self.name = name
         self.input_dim = input_dim
         self.hidden_dim = output_dim
         self.output_dim = output_dim
         self.input_layer = input_layer
-        self.X = input_layer.output().dimshuffle(1, 0, 2)
+        self.X = input_layer.output()
         self.s0 = s0
         self.params = []
 
@@ -127,7 +127,7 @@ class GRU(Layer):
             outputs_info=outputs_info
             )
 
-        self.Y = states.dimshuffle(1, 0, 2)
+        self.Y = states
 
     def output(self):
         return self.Y

theano/gpuarray/tests/test_dnn.py

@@ -1457,7 +1457,7 @@ def test_dnn_rnn_gru():
         numpy.zeros((psize,), dtype=theano.config.floatX))
 
     model = Model()
-    last_layer = WrapperLayer(X.dimshuffle(1, 0, 2))
+    last_layer = WrapperLayer(X)
     last_dim = input_dim
     for i in range(depth):
         gru = GRU(last_dim, hidden_dim, last_layer, s0=h0[i, :, :])
@@ -1478,7 +1478,7 @@ def test_dnn_rnn_gru():
         grad_fn = theano.function([X, Y, h0], grad, mode=mode_with_gpu)
         return fn, grad_fn
 
-    ref_fn, ref_grad_fn = funcs(last_layer.output().dimshuffle((1, 0, 2)),
+    ref_fn, ref_grad_fn = funcs(last_layer.output(),
                                 model.get_params())
     cudnn_fn, cudnn_grad_fn = funcs(rnnb.apply(params_cudnn, X, h0)[0],
                                     [params_cudnn])