!612 Bugfix: correct wrong pattern of confusion_softmax_grad_rule pass

Merge pull request !612 from huanghui/fix-confusion-softmax-grad-rule-pass

mindspore/ccsrc/pre_activate/ascend/ir_fusion/confusion_softmax_grad_rule.cc

@@ -47,7 +47,7 @@ void SetAttrsForFusionNode(const AnfNodePtr &sub_anf, const AnfNodePtr &fusion_n
 
 const BaseRef ConfusionSoftmaxGradRule::DefinePattern() const {
   return VectorRef(
-    {prim::kPrimSub, input0_, VectorRef({prim::kPrimReduceSum, VectorRef({prim::kPrimMul, input0_, input1_})})});
+    {prim::kPrimSub, input0_, VectorRef({prim::kPrimReduceSum, VectorRef({prim::kPrimMul, input1_, input0_})})});
 }
 
 const AnfNodePtr ConfusionSoftmaxGradRule::Process(const FuncGraphPtr &graph, const AnfNodePtr &node,

tests/ut/cpp/python_input/gtest_input/pre_activate/confusion_softmax_grad_rule.py

@@ -41,7 +41,7 @@ def test_confusion_softmax_grad_rule(tag):
 
     @fns
     def before(input0, input1):
-        res = mul(input0, input1)
+        res = mul(input1, input0)
         # input axis will be convert to attr in ConstructKernelGraph step
         res = reduce_sum(res, axis)
         res = sub(input0, res)