[Fix]: Fix deadlock issue related with MMDetWandbHook (#9476)

Co-authored-by: WangYudong <yudong.wang@akane.waseda.jp> Co-authored-by: WangYudong <yudong.wang@akane.waseda.jp>
open-mmlab · Dec 12, 2022 · cc9513f · cc9513f
1 parent 0dfe3f0
commit cc9513f
Showing 1 changed file with 5 additions and 1 deletion.
diff --git a/mmdet/core/hook/wandblogger_hook.py b/mmdet/core/hook/wandblogger_hook.py
@@ -235,7 +235,8 @@ def after_train_epoch(self, runner):
             # Log the table
             self._log_eval_table(runner.epoch + 1)
 
-    @master_only
+    # for the reason of this double-layered structure, refer to
+    # https://github.com/open-mmlab/mmdetection/issues/8145#issuecomment-1345343076
     def after_train_iter(self, runner):
         if self.get_mode(runner) == 'train':
             # An ugly patch. The iter-based eval hook will call the
@@ -245,7 +246,10 @@ def after_train_iter(self, runner):
             return super(MMDetWandbHook, self).after_train_iter(runner)
         else:
             super(MMDetWandbHook, self).after_train_iter(runner)
+        self._after_train_iter(runner)
 
+    @master_only
+    def _after_train_iter(self, runner):
         if self.by_epoch:
             return