[ST][MS][教程]resnet50图像分类在910A上graph模式训练精度达不到官网的0.74

name	about	labels
Bug Report	Use this template for reporting a bug	kind/bug

Describe the current behavior / 问题描述 (Mandatory / 必填)

resnet50图像学习在910A上graph模式训练精度达不到官网的0.74

Environment / 环境信息 (Mandatory / 必填)

Hardware Environment(Ascend/GPU/CPU) / 硬件环境:

Please delete the backend not involved / 请删除不涉及的后端:
/device ascend

Software Environment / 软件环境 (Mandatory / 必填):
-- MindSpore version (e.g., 1.7.0.Bxxx) :
-- Python version (e.g., Python 3.7.5) :
-- OS platform and distribution (e.g., Linux Ubuntu 16.04):
-- GCC/Compiler version (if compiled from source):
Excute Mode / 执行模式 (Mandatory / 必填)(PyNative/Graph):

Please delete the mode not involved / 请删除不涉及的模式:
/mode graph

Related testcase / 关联用例 (Mandatory / 必填)

用例路径：solution_test/cases/03subject_test/06document/02network_cases
关联用例：test_ms_tutorial_cv_resnet50_image_classification_0001.py

Steps to reproduce the issue / 重现步骤 (Mandatory / 必填)

进入官网，https://www.mindspore.cn/tutorials/application/zh-CN/master/cv/resnet50.html
下载样例代码，使用python执行
验证教程网络训练是否成功，日志中是否存在告警日志

Describe the expected behavior / 预期结果 (Mandatory / 必填)

网络训练成功，精度达到0.74

Related log / screenshot / 日志 / 截图 (Mandatory / 必填)

Epoch: [  1/  5], Steps: [  1/196], Train Loss: [2.312]
Epoch: [  1/  5], Steps: [101/196], Train Loss: [1.825]
Epoch: [  1/  5], Steps: [196/196], Train Loss: [1.425]
--------------------------------------------------
Epoch: [  1/  5], Average Train Loss: [1.873], Accuracy: [0.472]
--------------------------------------------------
Epoch: [  2/  5], Steps: [  1/196], Train Loss: [1.543]
Epoch: [  2/  5], Steps: [101/196], Train Loss: [1.393]
Epoch: [  2/  5], Steps: [196/196], Train Loss: [1.366]
--------------------------------------------------
Epoch: [  2/  5], Average Train Loss: [1.348], Accuracy: [0.564]
--------------------------------------------------
Epoch: [  3/  5], Steps: [  1/196], Train Loss: [1.150]
Epoch: [  3/  5], Steps: [101/196], Train Loss: [1.208]
Epoch: [  3/  5], Steps: [196/196], Train Loss: [1.011]
--------------------------------------------------
Epoch: [  3/  5], Average Train Loss: [1.175], Accuracy: [0.600]
--------------------------------------------------
Epoch: [  4/  5], Steps: [  1/196], Train Loss: [1.045]
Epoch: [  4/  5], Steps: [101/196], Train Loss: [1.151]
Epoch: [  4/  5], Steps: [196/196], Train Loss: [0.923]
--------------------------------------------------
Epoch: [  4/  5], Average Train Loss: [1.100], Accuracy: [0.617]
--------------------------------------------------
Epoch: [  5/  5], Steps: [  1/196], Train Loss: [0.969]
Epoch: [  5/  5], Steps: [101/196], Train Loss: [0.941]
Epoch: [  5/  5], Steps: [196/196], Train Loss: [1.063]
--------------------------------------------------
Epoch: [  5/  5], Average Train Loss: [1.069], Accuracy: [0.623]

Special notes for this issue/备注 (Optional / 选填)

走给吕昱峰

Please assign maintainer to check this issue.
请为此issue分配处理人。
@baimz

感谢您的提问，您可以评论//mindspore-assistant更快获取帮助：

如果您刚刚接触MindSpore，或许您可以在教程找到答案
如果您是资深Pytorch用户，您或许需要：

如果您遇到动态图问题，可以设置set_context(pynative_synchronize=True)查看报错栈协助定位
模型精度调优问题可参考官网调优指南
如果您反馈的是框架BUG，请确认您在ISSUE中提供了MindSpore版本、使用的后端类型（CPU、GPU、Ascend）、环境、训练的代码官方链接以及可以复现报错的代码的启动方式等必要的定位信息
如果您已经定位出问题根因，欢迎提交PR参与MindSpore开源社区，我们会尽快review

训练到11个epoch才达到官网的指标
输入图片说明

CCB结论：
遗留原因：910A从VM切换到GE后架构发生变动
风险：外部用户使用本教程时设为GRAPH_MODE时会碰到精度收敛慢的问题
规避措施：本版本遗留，630版本继续解决

二分法定位到commitID为0140d5d8eea76c016454fac1d779492dd84ca818导致出现的问题，相关PRhttps://gitee.com/mindspore/mindspore/pulls/67037

GVP MindSpore / mindspore

内容风险标识