Merge pull request #5288 from chinnadhurai/ccw_4916

theano/sandbox/rng_mrg.py

@@ -24,7 +24,8 @@ from . import multinomial
 
 import theano.sandbox.cuda
 from theano.sandbox.cuda import GpuOp
-from theano.gpuarray.basic_ops import GpuKernelBase, Kernel, infer_context_name
+from theano.sandbox.cuda.basic_ops import as_cuda_ndarray_variable
+from theano.gpuarray.basic_ops import GpuKernelBase, Kernel, infer_context_name, as_gpuarray_variable
 from theano.gpuarray.type import GpuArrayType
 from theano.gpuarray.fp16_help import write_w
 from theano.gpuarray.opt import (register_opt as register_gpua,
@@ -312,19 +313,6 @@ class mrg_uniform_base(Op):
             s = "no_inplace"
         return self.__class__.__name__ + "{%s,%s}" % (self.output_type, s)
 
-    def make_node(self, rstate, size):
-        # error checking slightly redundant here, since
-        # this op should not be called directly.
-        #
-        # call through MRG_RandomStreams instead.
-        broad = []
-        for i in range(self.output_type.ndim):
-                broad.append(tensor.extract_constant(size[i]) == 1)
-        output_type = self.output_type.clone(broadcastable=broad)()
-        return Apply(self,
-                     [rstate, size],
-                     [rstate.type(), output_type])
-
     def grad(self, inputs, ograd):
         return [gradient.grad_undefined(self, k, inp,
                                         'No gradient defined through '
@@ -338,6 +326,20 @@ class mrg_uniform_base(Op):
 class mrg_uniform(mrg_uniform_base):
     # CPU VERSION
 
+    def make_node(self, rstate, size):
+        # error checking slightly redundant here, since
+        # this op should not be called directly.
+        #
+        # call through MRG_RandomStreams instead.
+        broad = []
+        for i in range(self.output_type.ndim):
+                broad.append(tensor.extract_constant(size[i]) == 1)
+        output_type = self.output_type.clone(broadcastable=broad)()
+        rstate = as_tensor_variable(rstate)
+        return Apply(self,
+                     [rstate, size],
+                     [rstate.type(), output_type])
+
     @classmethod
     def new(cls, rstate, ndim, dtype, size):
         v_size = as_tensor_variable(size)
@@ -564,6 +566,20 @@ class mrg_uniform(mrg_uniform_base):
 class GPU_mrg_uniform(mrg_uniform_base, GpuOp):
     # GPU VERSION
 
+    def make_node(self, rstate, size):
+        # error checking slightly redundant here, since
+        # this op should not be called directly.
+        #
+        # call through MRG_RandomStreams instead.
+        broad = []
+        for i in range(self.output_type.ndim):
+                broad.append(tensor.extract_constant(size[i]) == 1)
+        output_type = self.output_type.clone(broadcastable=broad)()
+        rstate = as_cuda_ndarray_variable(rstate)
+        return Apply(self,
+                     [rstate, size],
+                     [rstate.type(), output_type])
+
     @classmethod
     def new(cls, rstate, ndim, dtype, size):
         v_size = as_tensor_variable(size)
@@ -809,6 +825,20 @@ class GPUA_mrg_uniform(GpuKernelBase, mrg_uniform_base):
     # GpuArray version
     _f16_ok = True
 
+    def make_node(self, rstate, size):
+        # error checking slightly redundant here, since
+        # this op should not be called directly.
+        #
+        # call through MRG_RandomStreams instead.
+        broad = []
+        for i in range(self.output_type.ndim):
+                broad.append(tensor.extract_constant(size[i]) == 1)
+        output_type = self.output_type.clone(broadcastable=broad)()
+        rstate = as_gpuarray_variable(rstate, infer_context_name(rstate))
+        return Apply(self,
+                     [rstate, size],
+                     [rstate.type(), output_type])
+
     def get_params(self, node):
         return node.inputs[0].type.context
 

theano/sandbox/tests/test_rng_mrg.py

@@ -18,6 +18,7 @@ from theano.sandbox.rng_mrg import MRG_RandomStreams
 from theano.sandbox.cuda import cuda_available
 from theano.tests import unittest_tools as utt
 from theano.tests.unittest_tools import attr
+import theano.gpuarray.tests.config
 
 if cuda_available:
     from theano.sandbox.cuda import float32_shared_constructor
@@ -1178,6 +1179,16 @@ def test_overflow_gpu_new_backend():
     rng_mrg_overflow(sizes, fct, mode, should_raise_error=False)
 
 
+def test_validate_input_types_gpuarray_backend():
+    from theano.sandbox.rng_mrg import mrg_uniform
+    from theano.gpuarray.type import gpuarray_shared_constructor
+    from theano.configparser import change_flags
+
+    with change_flags(compute_test_value="raise"):
+        rstate = numpy.zeros((7, 6), dtype="int32")
+        rstate = gpuarray_shared_constructor(rstate)
+        mrg_uniform.new(rstate, ndim=None, dtype="float32", size=(3,))
+
 if __name__ == "__main__":
     rng = MRG_RandomStreams(numpy.random.randint(2147462579))
     print(theano.__file__)