Constant対応に備えてリファクタ

python/src/nnabla/utils/converter/onnx/reader.py

@@ -52,7 +52,7 @@
 }
 
 
-def add_value_info_as_variable(info, network):
+def add_value_info_as_variable(network, info):
     if not info.type.HasField("tensor_type"):  # accepting only tensor
         raise ValueError("Only TensorProto is allowed as ValueInfoProto's type for info.name (Got {})"
                          .format(info.name, info.type))
@@ -62,15 +62,15 @@ def add_value_info_as_variable(info, network):
     v.shape.dim.extend([x.dim_value for x in t.shape.dim])
     return v
 
-def add_value_info_as_parameter(info, network):
-    v = add_value_info_as_variable(info, network)
+def add_value_info_as_parameter(network, info):
+    v = add_value_info_as_variable(network, info)
     v.type = "Parameter"
     v.initializer.type = "Constant"
     v.initializer.multiplier = 1.0
     return v
 
-def add_value_info_as_buffer(info, network):
-    v = add_value_info_as_variable(info, network)
+def add_value_info_as_buffer(network, info):
+    v = add_value_info_as_variable(network, info)
     v.type = "Buffer"
     return v
 
@@ -623,11 +623,11 @@ def onnx_graph_to_nnp_protobuf(pb, graph):
     for i in graph.input:
         if i.name in param_vars:
             # This input is a parameter
-            v = add_value_info_as_parameter(i, network)
+            v = add_value_info_as_parameter(network, i)
             param_list.append(v)
         else:
             # This input is a buffer
-            v = add_value_info_as_buffer(i, network)
+            v = add_value_info_as_buffer(network, i)
             in_list.append(v)
         if i.name in all_vars:
             del all_vars[i.name]
@@ -638,12 +638,17 @@ def onnx_graph_to_nnp_protobuf(pb, graph):
             # No one is using this buffer so we show a warning and ignore it.
             logger.warning("Input buffer {} is not used as input for any node.".format(i.name))
     for o in graph.output:
-        v = add_value_info_as_buffer(o, network)
+        v = add_value_info_as_buffer(network, o)
         out_list.append(v)
         del all_vars[v.name]
 
     for varg in all_vars:
-        # We add all remaining variables as intermediate buffer
+        # We add all remaining variables as intermediate buffer,
+        # except for the ones that was converted to a parameter.
+        # A conversion of a buffer to a parameter may occur in functions
+        # such as Constant
+        if varg in param_vars:
+            pass
         # We leave the buffer size of all intermediate buffers empty
         v = network.variable.add()
         v.type = "Buffer"