-
Notifications
You must be signed in to change notification settings - Fork 5.9k
Gpu samplecode test On PR-CPU-Py2 #33634
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Gpu samplecode test On PR-CPU-Py2 #33634
Conversation
|
Thanks for your contribution! |
0aa3d44 to
db8c523
Compare
9730b5c to
6d1db56
Compare
35f523d to
8122b5b
Compare
paddle/scripts/paddle_build.sh
Outdated
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.
此处解释: $spec_path 即 paddle/fluid/API_PR.spec 和 paddle/fluid/API_DEV.spec 文件, print_signatures.py 已更改为不打印ArgSpec,而且流水线环境中也没有py2环境均为py3环境。故删除此处逻辑
tools/check_api_approvals.sh
Outdated
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.
此处解释:
这两个程序已在 48行 62行 分别执行过。不需要再单独执行了。
另,下面的判断exit的代码应和上面的if判断 合并。
MingMingShangTian
left a comment
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.
LGTM
8122b5b to
3067008
Compare
3067008 to
7e5629f
Compare
7e5629f to
57a1610
Compare
57a1610 to
8e050ce
Compare
WITH_GPU == ON save
test=document_fix
test=document_fix
038ffc7 to
1fb196f
Compare
1fb196f to
bcf950c
Compare
test=document_fix save
cb1f6bb to
4324ec8
Compare
|
@wawltor 看一看,把 ops.py 的 |
pangyoki
left a comment
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.
LGTM
PR types
Function optimization
PR changes
Others
Describe
在PR-CPU-Py2流水线执行GPU示例代码测试,把流水线的版本改为GPU版本。
// 对啊,第一个怎么办
see http://agroup.baidu.com/paddlepaddle-org-cn/md/article/4036225#0618
已测三次,第一次运行时间5小时;第二次运行时间2.5小时,第三次运行时间接近3小时。
因为编译cache的有益影响,第二次编译非常快就完成了。但是整体增加时间确实太长了。等于将py2流水线增加了约两个小时。
而若在已有的py3流水线上执行个全量测试,增加时间也就10分钟左右(单卡P4上测得)。
06-21:经过多次运行之后,编译缓存也见效显著。编译代码未做变更时,编译时长仅为17分钟左右。与CPU的9分钟相比增加8分钟。
代码少量变更的roll optimize by SunNy820828449 · Pull Request #32880 · PaddlePaddle/Paddle任务时长为28分钟。可以接受。与之前的9-15分钟相比,增长<20分钟。
但有更改主体代码的仅此一例。需要合入此PR代码后继续观察。且此逻辑由
WITH_GPU直接影响,仅需修改PRC-CI-CPU-Py2的执行脚本内的环境变量即可控制本逻辑。0622
补充修改:http://agroup.baidu.com/paddlepaddle-org-cn/md/article/4036225
关于ops.py 中的
func变量,改为私有_func。因为func是个临时变量。 昨天的一个修改把这个导出 print signatures 打印出来了。这个临时变量 可能因为顺序每次不一样,然后每次就都与另一个不同的op的md5相同, 也有可能顺序正好一致 就通过了CI, 顺序不一致就通不过。