fix some bugs for some optimizer not call super init, test=allcases

PaddlePaddle · Aug 3, 2022 · 6b1ff3a · 6b1ff3a
1 parent 80e2c56
commit 6b1ff3a
Showing 1 changed file with 10 additions and 12 deletions.
diff --git a/python/paddle/optimizer/optimizer.py b/python/paddle/optimizer/optimizer.py
@@ -233,11 +233,6 @@ def __init__(self,
             else:
                 self._dtype = self._parameter_list[0].dtype
 
-        # learning rate can't be float16, for pure fp16 training, should extra handle the dtype for lr
-        self._lr_dtype = paddle.get_default_dtype(
-        ) if self._dtype is None else self._dtype
-        self._lr_dtype = paddle.float32 if self._lr_dtype == paddle.float16 else self._lr_dtype
-
         # each program should have a independent learning rate
         # program -> tensor(learning_rate)
         self._learning_rate_map = dict()
@@ -390,18 +385,21 @@ def get_opti_var_name_list(self):
         return self._opti_name_list
 
     def _create_global_learning_rate(self):
+        # learning rate can't be float16, for pure fp16 training, should extra handle the dtype for lr
+        _lr_dtype = paddle.get_default_dtype(
+        ) if self._dtype is None else self._dtype
+        _lr_dtype = paddle.float32 if self._lr_dtype == paddle.float16 else self._lr_dtype
         if isinstance(self._learning_rate, LRScheduler):
             lr_var = self._global_learning_rate()
             # only create global lr_var once
             if not isinstance(lr_var, framework.Variable):
                 lr_name = unique_name.generate('learning_rate')
                 self._learning_rate._var_name = lr_name
-                lr_var = self.helper.create_global_variable(
-                    name=lr_name,
-                    shape=[1],
-                    persistable=True,
-                    stop_gradient=True,
-                    dtype=self._lr_dtype)
+                lr_var = self.helper.create_global_variable(name=lr_name,
+                                                            shape=[1],
+                                                            persistable=True,
+                                                            stop_gradient=True,
+                                                            dtype=_lr_dtype)
                 main_prog = framework.default_main_program()
                 main_prog.lr_sheduler = self._learning_rate
                 main_prog.lr_var = lr_var
@@ -423,7 +421,7 @@ def _create_global_learning_rate(self):
                     name=unique_name.generate("learning_rate"),
                     shape=[1],
                     value=float(self._learning_rate),
-                    dtype=self._lr_dtype,
+                    dtype=_lr_dtype,
                     persistable=True)
 
     @framework.dygraph_only