fix training hang in PaddleCloud single machine. (#1233)

cf180c1f · Kaipeng Deng · GitHub · 1cc363f6 · cf180c1f
显示空白变更内容
内联并排

Showing with 3 addition and 1 deletion

tools/train.py tools/train.py +3 -1

未找到文件。
--- a/tools/train.py
+++ b/tools/train.py
@@ -53,7 +53,9 @@ logger = logging.getLogger(__name__)

 def main():
    env = os.environ
-    FLAGS.dist = 'PADDLE_TRAINER_ID' in env and 'PADDLE_TRAINERS_NUM' in env
+    FLAGS.dist = 'PADDLE_TRAINER_ID' in env \
+                    and 'PADDLE_TRAINERS_NUM' in env \
+                    and int(env['PADDLE_TRAINERS_NUM']) > 1
    if FLAGS.dist:
        trainer_id = int(env['PADDLE_TRAINER_ID'])
        local_seed = (99 + trainer_id)