-
Notifications
You must be signed in to change notification settings - Fork 39
Conversation
It seems that in AWS, CPU performance has some stable difference.
add multi card for text_classification
Update actived to false for image classification
Please do NOT merge, PR is not ready yet. |
PR is ready for review |
陈老师, 你这个pr有点大啊,看你这个ce task check已经过了, 应该问题不大。 |
我看了下ce task的job, 这个task 总共大概1小时, 其中编译+ 部署大概40min, 这个时间略长,后续考虑怎么优化下吧 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
现在可以直接�向 master 提pr,会有 ci 测
可以加下description 吗,不清楚在做啥
vgg16_aws_dist/ce_runner.py
Outdated
kpi_to_track = LessWorseKpi(name, 0.01) | ||
kpi_to_track.add_record(np.array(val, dtype='float32')) | ||
|
||
class DataCollector(object): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please run pre-commit run -a
to automatically format the code style.
this PR is rebased and recommitted with another one #33 |
@guochaorong 编译主要是CE在编译最新的paddle, 这个是我这边没有控制的. 编译完成后生成docker image并且push出去应该不会太久, 剩下就是创建aws instance和执行测试, 这个过程是这个测试里面耗时最长的, 但是也没有什么可以优化的了, 可以并行的都已经在并行测试了. 目前是安排了三个cluster config的测试, 相当于普通测试要跑1.5遍(前两个并行, 最后一个串行), 并且要加上instance的allocate时间. |
@Superjomn 重新rebase后往master提交了新的PR. 增加了一些注释来解释过程. 提交前执行了pre-commit. |
No description provided.