[cherry pick] Modify adam to adamw in Optimizer AdamW (#36028) (#36103)

The AdamW optimizer modify the op from adamw to adam in pr35521, this is a inappropriate modify. Modify adam to adamw in AdamW.

Author: zhangbo9674

paddle/fluid/pybind/op_function_generator.cc

@@ -71,6 +71,9 @@ std::map<std::string, std::set<std::string>> op_ins_map = {
     {"adam",
      {"Param", "Grad", "LearningRate", "Moment1", "Moment2", "Beta1Pow",
       "Beta2Pow", "MasterParam"}},
+    {"adamw",
+     {"Param", "Grad", "LearningRate", "Moment1", "Moment2", "Beta1Pow",
+      "Beta2Pow", "MasterParam"}},
 };
 
 // NOTE(zhiqiu): Like op_ins_map.
@@ -110,6 +113,9 @@ std::map<std::string, std::set<std::string>> op_outs_map = {
     {"adam",
      {"ParamOut", "Moment1Out", "Moment2Out", "Beta1PowOut", "Beta2PowOut",
       "MasterParamOut"}},
+    {"adamw",
+     {"ParamOut", "Moment1Out", "Moment2Out", "Beta1PowOut", "Beta2PowOut",
+      "MasterParamOut"}},
 };
 
 // NOTE(zhiqiu): Commonly, the outputs in auto-generated OP function are
@@ -129,7 +135,8 @@ std::map<std::string, std::set<std::string>> op_passing_outs_map = {
      {"ParamOut", "Moment1Out", "Moment2Out", "Beta1PowOut", "Beta2PowOut",
       "MasterParamOut"}},
     {"adamw",
-     {"ParamOut", "Moment1Out", "Moment2Out", "Beta1PowOut", "Beta2PowOut"}},
+     {"ParamOut", "Moment1Out", "Moment2Out", "Beta1PowOut", "Beta2PowOut",
+      "MasterParamOut"}},
     {"average_accumulates",
      {"out_sum_1", "out_sum_2", "out_sum_3", "out_num_accumulates",
       "out_old_num_accumulates", "out_num_updates"}},

python/paddle/optimizer/adamw.py

@@ -298,14 +298,14 @@ def _append_optimize_op(self, block, param_and_grad):
             _beta2 = self._beta2 if not isinstance(
                 self._beta2, Variable) else self._beta2.numpy().item(0)
 
-            _, _, _, _, _, _ = _C_ops.adam(
+            _, _, _, _, _, _ = _C_ops.adamw(
                 param_and_grad[0], param_and_grad[1], lr, moment1, moment2,
                 beta1_pow_acc, beta2_pow_acc, master_weight, param_and_grad[0],
                 moment1, moment2, beta1_pow_acc, beta2_pow_acc, master_weight,
                 'epsilon', self._epsilon, 'lazy_mode', self._lazy_mode,
                 'min_row_size_to_use_multithread', 1000, 'beta1', _beta1,
                 'beta2', _beta2, 'coeff', self._coeff, 'multi_precision',
-                find_master)
+                find_master, "lr_ratio", lr_ratio_)
 
             return None