more test cases

python/mxnet/contrib/onnx/mx2onnx/_op_translations.py

@@ -744,7 +744,7 @@ def convert_blockgrad(node, **kwargs):
     """ Skip operator  """
     return create_basic_op_node('ConstantFill', node, kwargs)
 
-@mx_op.register("make_loss")
+@mx_op.register("MakeLoss")
 def convert_makeloss(node, **kwargs):
     """ Skip operator  """
     return create_basic_op_node('ConstantFill', node, kwargs)

python/mxnet/contrib/onnx/onnx2mx/_translation_utils.py

@@ -158,7 +158,7 @@ def _fix_broadcast(op_name, inputs, broadcast_axis, proto_obj):
         assert len(list(inputs)) == 2
 
         input0_shape = get_input_shape(inputs[0], proto_obj)
-        #creating reshape shape
+        # creating reshape shape
         reshape_shape = list(len(input0_shape) * (1,))
         reshape_shape[broadcast_axis] = -1
         reshape_shape = tuple(reshape_shape)

tests/python-pytest/onnx/export/mxnet_export_test.py

@@ -330,6 +330,27 @@ def test_scalarops():
             out = ipsym / 2
             _test_scalar_op(input1, out, np_out)
 
+def test_makeloss():
+    v1 = mx.nd.array([1, 2])
+    v2 = mx.nd.array([0, 1])
+    a = mx.sym.Variable('a')
+    b = mx.sym.Variable('b')
+    sym = mx.sym.MakeLoss(b + a)
+    ex = sym.bind(ctx=mx.cpu(0), args={'a': v1, 'b': v2})
+    ex.forward(is_train=True)
+    makeloss_out = ex.outputs[0].asnumpy()
+
+    converted_model = onnx_mxnet.export_model(sym, {}, [(2,),(2,)], np.float32, "makelossop.onnx")
+
+    # sym, arg_params, aux_params = onnx_mxnet.import_model(converted_model)
+    # result = forward_pass(sym, arg_params, aux_params, ['ipsym'], input1)
+    #
+    # # Comparing result of forward pass before using onnx export, import
+    # npt.assert_almost_equal(result, makeloss_out)
+    #
+    # executor = loss.simple_bind(ctx=mx.cpu(0), a=(2,), b=(2,))
+    # executor.forward(is_train=True, a=v1, b=v2)
+    # print(executor.outputs[0])
 
 @with_seed()
 def test_comparison_ops():