add onnx scalar elt

mace/python/tools/converter_tool/onnx_converter.py

@@ -941,6 +941,33 @@ class OnnxConverter(base_converter.ConverterInterface):
             coeff_arg = op.arg.add()
             coeff_arg.name = MaceKeyword.mace_coeff_str
             coeff_arg.floats.extend([min_value, max_value])
+        elif len(node.inputs) == 2:
+            if node.inputs[1] in self._consts and \
+                    node.inputs[0] not in self._consts:
+                const_name = node.inputs[1]
+                const_tensor = self._consts[const_name]
+                if len(const_tensor.dims) == 0:
+                    value_arg = op.arg.add()
+                    value_arg.name = MaceKeyword.mace_scalar_input_str
+                    value_arg.f = const_tensor.float_data[0]
+                    value_index_arg = op.arg.add()
+                    value_index_arg.name = \
+                        MaceKeyword.mace_scalar_input_index_str
+                    value_index_arg.i = 1
+                    del op.input[1]
+            elif node.inputs[0] in self._consts and \
+                    node.inputs[1] not in self._consts:
+                const_name = node.inputs[0]
+                const_tensor = self._consts[const_name]
+                if len(const_tensor.dims) == 0:
+                    value_arg = op.arg.add()
+                    value_arg.name = MaceKeyword.mace_scalar_input_str
+                    value_arg.f = const_tensor.float_data[0]
+                    value_index_arg = op.arg.add()
+                    value_index_arg.name = \
+                        MaceKeyword.mace_scalar_input_index_str
+                    value_index_arg.i = 0
+                    del op.input[0]
 
     @staticmethod
     def copy_node_attr(op, node, attr_name, dtype=AttributeType.INT,

tools/onnx_optimizer.py

@@ -22,12 +22,12 @@ from onnx import optimizer
 
 def main():
     if len(sys.argv) != 3:
-        print "Usage: python onnx_optimizer.py model.onnx model_opt.onnx"
+        print("Usage: python onnx_optimizer.py model.onnx model_opt.onnx")
         sys.exit(0)
     in_path = sys.argv[1]
     out_path = sys.argv[2]
     original_model = onnx.load(in_path)
-    print "Start optimize ONNX model for inference:"
+    print("Start optimize ONNX model for inference:")
     passes = ['eliminate_identity',
               'fuse_consecutive_squeezes',
               'fuse_consecutive_transposes',
@@ -35,15 +35,14 @@ def main():
               'eliminate_nop_transpose',
               'eliminate_unused_initializer',
               'extract_constant_to_initializer',
-              'fuse_add_bias_into_conv',
               'fuse_bn_into_conv',
               'fuse_transpose_into_gemm']
     for i in range(len(passes)):
-        print i, ".", passes[i]
+        print("%s.%s" % (i, passes[i]))
     optimized_model = optimizer.optimize(original_model, passes)
     onnx.save_model(optimized_model, out_path)
-    print "Optimize Finished!"
-    print "Please check new model in:", out_path
+    print("Optimize Finished!")
+    print("Please check new model in:", out_path)
 
 
 if __name__ == '__main__':