ONNXのhelperを使うよう修正

python/test/utils/test_conversion.py

@@ -18,6 +18,7 @@
 import nnabla.utils.load as nnload
 from nnabla.utils import nnabla_pb2
 import onnx
+import onnx.helper
 from onnx import (ModelProto, TensorProto, GraphProto,
     TensorShapeProto, AttributeProto, NodeProto)
 import nnabla.logger as logger
@@ -343,51 +344,37 @@ def read(self):
         return onnx_model_to_nnp_protobuf(model_proto)
 
 def convert_to_node(func, variables):
-    n = NodeProto()
-    n.name = func.name
-    n.op_type = nnabla_function_type_to_onnx_optype.get(func.type, func.type)
-    n.input.extend(func.input)
-    n.output.extend(func.output)
+    n = onnx.helper.make_node(
+            nnabla_function_type_to_onnx_optype.get(func.type, func.type),
+            func.input,
+            func.output,
+            name=func.name)
     if func.type == "Concatenate":
         # ONNX requires axis setting as a parameter
         # for the concat op_type.
-        attr = n.attribute.add()
-        attr.name = "axis"
-        attr.type = AttributeProto.INT
         # If no value is set for axis,
         # the default value 0 will be set
-        attr.i = func.concatenate_param.axis
+        attr = onnx.helper.make_attribute("axis", func.concatenate_param.axis)
+        n.attribute.extend([attr])
     elif func.type == "Dropout":
         # NNP Dropout is always is_test=false
         # since we always apply dropout when it is
         # included in a network.
-        attr = n.attribute.add()
-        attr.name = "is_test"
-        attr.type = AttributeProto.INT
-        attr.i = 0
+        attr = onnx.helper.make_attribute("is_test", 0)
+        n.attribute.extend([attr])
     elif func.type == "Identity":
         # Convert Identity to a Dropout with is_test=true
         # so we just copy the input to output
         n.op_type = "Dropout"
-        attr = n.attribute.add()
-        attr.name = "is_test"
-        attr.type = AttributeProto.INT
-        attr.i = 1
+        attr = onnx.helper.make_attribute("is_test", 1)
+        n.attribute.extend([attr])
     elif func.type == "MaxPooling":
         mpp = func.max_pooling_param
         # Copy kernel, stride, and pads values
-        k = n.attribute.add()
-        k.name = "kernel_shape"
-        k.type = AttributeProto.INTS
-        k.ints.extend(mpp.kernel.dim)
-        s = n.attribute.add()
-        s.name = "strides"
-        s.type = AttributeProto.INTS
-        s.ints.extend(mpp.stride.dim)
-        p = n.attribute.add()
-        p.name = "pads"
-        p.type = AttributeProto.INTS
-        p.ints.extend(mpp.pad.dim)
+        k = onnx.helper.make_attribute("kernel_shape", mpp.kernel.dim)
+        s = onnx.helper.make_attribute("strides", mpp.stride.dim)
+        p = onnx.helper.make_attribute("pads", mpp.pad.dim)
+        n.attribute.extend([k, s, p])
     elif func.type == "Convolution":
         cp = func.convolution_param
         # Calculate the kernel_shape from input weight data.
@@ -405,27 +392,12 @@ def convert_to_node(func, variables):
         weight_shape = weight_var[0].shape
         # The base axis for weights is the next axis from the data's base axis
         weight_base = cp.base_axis + 1
-        k = n.attribute.add()
-        k.name = "kernel_shape"
-        k.type = AttributeProto.INTS
-        k.ints.extend(weight_shape.dim[weight_base:])
-
-        d = n.attribute.add()
-        d.name = "dilations"
-        d.type = AttributeProto.INTS
-        d.ints.extend(cp.dilation.dim)
-        s = n.attribute.add()
-        s.name = "strides"
-        s.type = AttributeProto.INTS
-        s.ints.extend(cp.stride.dim)
-        p = n.attribute.add()
-        p.name = "pads"
-        p.type = AttributeProto.INTS
-        p.ints.extend(cp.pad.dim)
-        g = n.attribute.add()
-        g.name = "group"
-        g.type = AttributeProto.INT
-        g.i = cp.group
+        k = onnx.helper.make_attribute("kernel_shape", weight_shape.dim[weight_base:])
+        d = onnx.helper.make_attribute("dilations", cp.dilation.dim)
+        s = onnx.helper.make_attribute("strides", cp.stride.dim)
+        p = onnx.helper.make_attribute("pads", cp.pad.dim)
+        g = onnx.helper.make_attribute("group", cp.group)
+        n.attribute.extend([k, d, s, p, g])
     return n
 
 def nnp_model_to_onnx_graph(graph, nnp):
@@ -693,8 +665,7 @@ def change_to_copy(node):
 
     c2out = onnx_caffe2.backend.run_model(model, [img])
     # Process onnx with naabla
-    r = OnnxReader(path)
-    nnp = r.read()
+    nnp = onnx_model_to_nnp_protobuf(model)
     assert nnp is not None
     assert len(nnp.other_files) == 0
     assert nnp.protobuf is not None
@@ -705,9 +676,9 @@ def change_to_copy(node):
     nnpdir = tmpdir.mkdir("nnp")
     p = os.path.join(str(nnpdir), nnp_name)
     nnpex.export_nnp(p)
-    pdb.set_trace()
+    #pdb.set_trace()
     # read exported nnp and run network
-    nn_net = nnload.load([p])
+    #nn_net = nnload.load([p])
     #exe = run_executor(nn_net, exec_name)
     ##in_data = exe.variables["in_data_0"]
     ##print(in_data.variable_instance.d)