报错：
/home/disk1/task_data/history/20190916/6.app-user-20190916110604-8973--lizhiqing_core2_middlepage_rank_para_201909161105_paddlecloud/logs/workspace
Mon Sep 16 11:23:16 2019[1,9]:I0916 11:23:16.774300 21861 rpc_client.h:101] init rpc client with trainer_id 9
Mon Sep 16 11:23:21 2019[1,9]:Traceback (most recent call last):
Mon Sep 16 11:23:21 2019[1,9]: File "train.py", line 221, in
Mon Sep 16 11:23:21 2019[1,9]: train_main(use_cuda=use_cuda, is_sparse=True, is_local=is_local)
Mon Sep 16 11:23:21 2019[1,9]: File "train.py", line 215, in train_main
Mon Sep 16 11:23:21 2019[1,9]: train_loop(t.get_trainer_program())
Mon Sep 16 11:23:21 2019[1,9]: File "train.py", line 185, in train_loop
Mon Sep 16 11:23:21 2019[1,9]: fetch_list=[avg_cost])
Mon Sep 16 11:23:21 2019[1,9]: File "/home/disk1/task_data/history/20190916/9.app-user-20190916110604-8973--lizhiqing_core2_middlepage_rank_para_201909161105_paddlecloud/logs/workspace/python27-gcc482/lib/python2.7/site-packages/paddle/fluid/executor.py", line 650, in run
Mon Sep 16 11:23:21 2019[1,9]: use_program_cache=use_program_cache)
Mon Sep 16 11:23:21 2019[1,9]: File "/home/disk1/task_data/history/20190916/9.app-user-20190916110604-8973--lizhiqing_core2_middlepage_rank_para_201909161105_paddlecloud/logs/workspace/python27-gcc482/lib/python2.7/site-packages/paddle/fluid/executor.py", line 748, in _run
Mon Sep 16 11:23:21 2019[1,9]: exe.run(program.desc, scope, 0, True, True, fetch_var_name)
Mon Sep 16 11:23:21 2019[1,9]:paddle.fluid.core_avx.EnforceNotMet: Invoke operator data_norm error.
相关代码：
2条答案
按热度按时间ttcibm8c1#
看组网的话没有明显错误，能否在本地复现？或者能否改用batch norm?
ep6jt1vc2#
嗯 好了 已经改用 batch_norm 发件人: Yi Liu notifications@github.com 答复: PaddlePaddle/Paddle reply@reply.github.com 日期: 2019年9月17日 星期二 上午9:49 收件人: PaddlePaddle/Paddle Paddle@noreply.github.com 抄送: "Li,Zhiqing" lizhiqing@baidu.com, Author author@noreply.github.com 主题: Re: [PaddlePaddle/Paddle] fluid1.5 在paddleclould 上运行报 data_norm error (#19812) 看组网的话没有明显错误，能否在本地复现？或者能否改用batch norm? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#19812?email_source=notifications&email_token=ALMGKM2K2CXTF7N24SWIKW3QKAZSBA5CNFSM4IW5KOUKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD63ATFQ#issuecomment-532023702>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ALMGKMY2SFZBIN3JJ2AKJNDQKAZSBANCNFSM4IW5KOUA>.