figure out dtype automatically in tensor_random

tensor_random.py

@@ -128,8 +128,12 @@ class NumpyGenerator(gof.op.Op):
         shape = tensor.convert_to_int64(_shape)
         if shape.type.ndim != 1:
             raise TypeError('shape argument was not converted to 1-d tensor', _shape)
+
+        # we generate one random number with the distribution to determine what dtype to expect
+        output_dtype = str(self.fn(numpy.random.RandomState(18), size=(1,)).dtype)
+
         inputs = [gof.Value(gof.type.generic, numpy.random.RandomState(self.seed)), shape]
-        outputs = [tensor.Tensor(dtype='float64', broadcastable = [False]*self.ndim).make_result()]
+        outputs = [tensor.Tensor(dtype=output_dtype, broadcastable = [False]*self.ndim).make_result()]
         return gof.Apply(op = self, inputs = inputs, outputs = outputs)
 
     def grad(self, inputs, grad_outputs):