[DoubleGrad PaddlePaddle#4] Bug Fixes to Double Grad Node Generation (PaddlePaddle#41121)

* [Refactor] refactored eager_gen.py PR #2

* [DoubleGrad PR #1] Decoupled code generation logics for Dygraph ForwardFunctions and GradNodes

* Fixed minor issue

* Adjusted logics of GenerateNodeCreationCodes and GenerateForwardDefinition

* Fixed issues

* Supported higher-order grad node generation

* [DoubleGrad PR PaddlePaddle#4] Supported higher-order GradNode generation

* [DoubleGrad PaddlePaddle#4] Bug Fixes to Double Grad Node Generation

* Fixed yaml typo

* Fixed yaml typo

* fixed minor issues

* Fixed minor issue

Author: jim19930609

paddle/fluid/eager/auto_code_generator/final_state_generator/codegen_utils.py

@@ -21,7 +21,8 @@
 ########################
 ### Global Variables ###
 ########################
-ops_to_fill_zero_for_empty_grads = set(["split", "rnn"])
+ops_to_fill_zero_for_empty_grads = set(
+    ["split_grad", "rnn_grad", "matmul_double_grad"])
 
 # For API dispatch used at python-level
 # { op_name : [arg_name, ...] }
@@ -176,6 +177,11 @@ def TransformGradVarNameForDoubleGradGeneration(string):
     return string
 
 
+def GetIndent(num):
+    tab = "   "
+    return "".join([tab for i in range(num)])
+
+
 ######################
 ###  Yaml Parsers  ###
 ######################