add dropout op map

paddle/phi/api/yaml/op_compat.yaml

@@ -414,6 +414,17 @@
 
 - op : dropout
   backward : dropout_grad
+  inputs :
+    x : X
+  outputs :
+    out : Out
+    mask : Mask
+  attrs :
+    p : dropout_prob
+    is_test : is_test
+    mode : dropout_implementation
+    seed : seed
+    fix_seed : fix_seed
   extra :
     attrs : [bool fix_seed = false, int seed = 0]
 
@@ -790,6 +801,14 @@
 
 - op : layer_norm
   backward : layer_norm_grad
+  inputs :
+    x : X
+    scale : Scale
+    bias : Bias
+  outputs :
+    out : Y
+    mean : Mean
+    variance : Variance
   extra :
     attrs : [bool use_mkldnn = false, str mkldnn_data_type = "float32", bool is_test = false]
 
@@ -940,6 +959,17 @@
   outputs :
     out : Out
 
+- op : mean (reduce_mean)
+  backward : reduce_mean_grad
+  inputs :
+    x : X
+  outputs :
+    out : Out
+  attrs :
+    {axis : dim, keepdim : keep_dim}
+  extra :
+    attrs : [bool use_mkldnn = false]
+
 - op : meshgrid
   backward : meshgrid_grad
   inputs :
@@ -1145,17 +1175,6 @@
   extra :
     attrs : [bool use_mkldnn = false]
 
-- op : mean (reduce_mean)
-  backward : reduce_mean_grad
-  inputs :
-    x : X
-  outputs :
-    out : Out
-  attrs :
-    {axis : dim, keepdim : keep_dim}
-  extra :
-    attrs : [bool use_mkldnn = false]
-
 - op : reduce_min
   backward : reduce_min_grad
   extra :

python/paddle/incubate/autograd/primx.py

@@ -597,11 +597,13 @@ def _lower_composite(block, blacklist=[]):
         # if output var of composite rule is None, this means this var is not needed
         none_vars_to_remove = set()
 
+        change = None
         # Step2: Process all ops in the target block
         for op_idx in range(len(block.ops)):
             op = block.ops[op_idx]
             ops_to_remove.append(op_idx)
             if lookup_fn(op.type) is not None and op.type not in blacklist:
+                change = True
                 input_args = prepare_python_api_arguments(op)
                 bind(input_args, to_bind, value_table)
 
@@ -681,6 +683,10 @@ def _lower_composite(block, blacklist=[]):
             block.desc._remove_var(var_name.encode())
             del block.vars[var_name]
         block._sync_with_cpp()
+
+        # composite ops may contain other ops, thus, call _lower_composite again.
+        if change:
+            _lower_composite(block, blacklist)
         return
 
     elif isinstance(block, typing.Sequence):

python/paddle/incubate/autograd/utils.py

@@ -232,14 +232,20 @@ def get_output_vars_from_comosite(op):
                 origin_output_name = op_map[name]["outputs"][item]
                 if origin_output_name not in origin_output_names:
                     continue
-                origin_output_var = get_var_block(op.block, op.output(origin_output_name))
+                origin_output_var = get_var_block(
+                    op.block, op.output(origin_output_name)
+                )
                 res.append(origin_output_var)
         elif len(origin_output_names) == 1:
             # When origin output num is 1, map info is not needed.
-            origin_output_var = get_var_block(op.block, op.output(origin_output_names[0]))
+            origin_output_var = get_var_block(
+                op.block, op.output(origin_output_names[0])
+            )
             res.append(origin_output_var)
         else:
-            raise ValueError("When replace op with composite rule, there must exist output map info from origin op to composite rule.")
+            raise ValueError(
+                "When replace op with composite rule, there must exist output map info from origin op to composite rule."
+            )
         return res