Change as graph.input can be either ValueInfoProto or string (#186) (#192)

* graph.input can be either ValueInfoProto or string

* pylint

nnvm/python/nnvm/frontend/onnx.py

@@ -245,18 +245,24 @@ class GraphProto(object):
                 raise ValueError("Tensor's name is required.")
             self._params[init_tensor.name] = self._parse_array(init_tensor)
         for i in graph.input:
-            if i in self._params:
+            # from onnx v0.2, GraphProto.input has type ValueInfoProto,
+            #  and the name is 'i.name'
+            try:
+                i_name = i.name
+            except AttributeError:
+                i_name = i
+            if i_name in self._params:
                 # i is a param instead of input
                 name_param = 'param_{}'.format(self._num_param)
                 self._num_param += 1
-                self._params[name_param] = self._params.pop(i)
+                self._params[name_param] = self._params.pop(i_name)
                 self._nodes[name_param] = _sym.Variable(name=name_param)
-                self._renames[i] = name_param
+                self._renames[i_name] = name_param
             else:
                 name_input = 'input_{}'.format(self._num_input)
                 self._num_input += 1
                 self._nodes[name_input] = _sym.Variable(name=name_input)
-                self._renames[i] = name_input
+                self._renames[i_name] = name_input
         # construct nodes, nodes are stored as directed acyclic graph
         for idx, node in enumerate(graph.node):
             op_name = node.op_type