Use coordinate_transformation_mode as default parameter (tf 2.0 will …

…change it) (#38)

onnxconverter_common/onnx_ops.py

@@ -571,7 +571,7 @@ def apply_reshape(scope, input_name, output_name, container, operator_name=None,
         container.add_node('Reshape', input_name, output_name, op_version=5, name=name)
 
 
-def apply_resize(scope, input_name, output_name, container, operator_name=None, mode='nearest', scales=None):
+def apply_resize(scope, input_name, output_name, container, operator_name=None, mode='nearest', coordinate_transformation_mode='asymmetric', scales=None):
     '''
     :param mode: "nearest" or "linear"
     :param scales: a float tensor for scaling (upsampling or downsampling) all input dimensions
@@ -590,7 +590,7 @@ def apply_resize(scope, input_name, output_name, container, operator_name=None,
         roi = [0.0] * len(scales) + [1.0] * len(scales)
         container.add_initializer(roi_tensor_name, onnx_proto.TensorProto.FLOAT, [2 * len(scales)], roi)
         inputs.append(roi_tensor_name)
-        attrs['coordinate_transformation_mode'] = 'asymmetric'
+        attrs['coordinate_transformation_mode'] = coordinate_transformation_mode
         if attrs['mode'] == 'nearest':
             attrs['nearest_mode'] = 'floor'
 
@@ -806,7 +806,7 @@ def apply_transpose(scope, input_name, output_name, container, operator_name=Non
     container.add_node('Transpose', input_name, output_name, name=name, perm=perm)
 
 
-def apply_upsample(scope, input_name, output_name, container, operator_name=None, mode='nearest', scales=None):
+def apply_upsample(scope, input_name, output_name, container, operator_name=None, mode='nearest', coordinate_transformation_mode='asymmetric', scales=None):
     '''
     :param mode: nearest or linear
     :param scales: an integer list of scaling-up rate of all input dimensions
@@ -838,4 +838,4 @@ def apply_upsample(scope, input_name, output_name, container, operator_name=None
     else:
         # Upsample op is deprecated in ONNX opset 10
         # We implement Upsample through Resize instead
-        apply_resize(scope, input_name, output_name, container, operator_name, mode, scales)
+        apply_resize(scope, input_name, output_name, container, operator_name, mode, coordinate_transformation_mode, scales)