Skip to content

93e4d0cce 相对 45c81a414f SgdSparseCpuTraining 慢了很多 #366

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.

Sign up for GitHub

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

Jump to bottom
Closed
linrongyi opened this issue Nov 5, 2016 · 5 comments
Closed

93e4d0cce 相对 45c81a414f SgdSparseCpuTraining 慢了很多 #366

linrongyi opened this issue Nov 5, 2016 · 5 comments
Assignees
@backyes

Comments

@linrongyi
Copy link

今天编译的时候checkout了最新的 93e4d0c, 发现SgdSparseCpuTraining 速度相对 45c81a4 慢了很多, 将近有10倍以上.

由于 @backyes 内部工具里面的mkl 工具没有 mkl_lapacke.h, 导致 93e4d0cce编译不过. 所以找了一个带有 mkl_lapacke.h 的MKL包. 两个版本都基于这个MKL编译的.

由于还没有修复GPU版本没有开启avx的功能, 所以编译的时候-DWITH_GPU=OFF.
@wangkuiyi
Copy link
Collaborator

wangkuiyi commented Nov 6, 2016 •
edited
Loading

@linrongyi 谢谢反馈这个问题。可否贴一下测试SgdSparseCpuTraining的速度的步骤？我们请paddle团队的同学也来复现一下这个问题，然后进一步找原因和解法。

@backyes backyes mentioned this issue Nov 6, 2016
Merged
@backyes
Copy link
Contributor

backyes commented Nov 6, 2016 •
edited
Loading

@linrongyi 我们在确认#359 的PR的意义。 这种10倍量级的性能差异极可能是使用上的问题，等我们确认mkl_lapacke.h的作用后再讨论解决方法。
@lzhao4ever 请知晓。

相关问题：
GPU avx BUG在 @hedaoyuan 的PR中#239 中已经FIX， 该BUG比较紧急，请 @hedaoyuan 看看能否提供一个单独FIX的patch。

@emailweixu
Copy link
Collaborator

@backyes 在mkl里加上mkl_lapacke.h即可

@luotao1
Copy link
Contributor

luotao1 commented Nov 7, 2016

In composer_xe_2013.0.079/mkl/include/mkl.h, it includes mkl_lapack.h:
mkl

@backyes backyes closed this as completed Nov 7, 2016
@backyes backyes reopened this Nov 7, 2016
@backyes
Copy link
Contributor

backyes commented Nov 7, 2016

@linrongyi 问题已经解决：

  • 性能问题。 根源是 @linrongyi 同学引用了一个性能存在问题的第三方MKL库
  • mkl_lapacke.h引入的编译问题。已经通过添加『mkl_lapacke.h』头文件，解决。（我们提供的MKL缺少相关库头文件，不缺少函数实现）

@backyes backyes closed this as completed Nov 7, 2016
wangxicoding pushed a commit to wangxicoding/Paddle that referenced this issue Dec 9, 2021
@Steffy-zxf
rm models module (PaddlePaddle#366)
fa6b699 
* rm models module
AnnaTrainingG pushed a commit to AnnaTrainingG/Paddle that referenced this issue Sep 19, 2022
@lzzyzlbb
add stargan pretrain model (PaddlePaddle#366)
f930083
WAYKEN-TSE pushed a commit to WAYKEN-TSE/Paddle that referenced this issue Dec 6, 2024
@NKNaN
AudioLDM2模型复现前向推理 (PaddlePaddle#366)
f049e2d 
任务：PaddlePaddle/PaddleMIX#250

- text-to-audio推理已跑通
zhangyuqin1998 pushed a commit to zhangyuqin1998/Paddle that referenced this issue Feb 20, 2025
@HouQiming @hwu36
Removed trivial copy constructors on parameter classes to enable devi… (
96a11a1 
PaddlePaddle#366)

* Removed trivial copy constructors on parameter classes to enable device-side launch of CUTLASS kernels

* Added SFINAE to the `TensorRef(NonConstTensorRef const&)` constructor to avoid making it a copy-constructor for device code

* std => platform

* fix affine2

* really fix affine2

Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
lizexu123 pushed a commit to lizexu123/Paddle that referenced this issue Mar 26, 2025
@qili93
Merge pull request PaddlePaddle#366 from graphcore/yzx-0804-mv
18e9c48 
[IPU] update docs and custom ops demo
tianyuzhou668 pushed a commit to tianyuzhou668/Paddle that referenced this issue May 12, 2025
@ronny1996
[Zero-Dim] add median, where 0-dim ut (PaddlePaddle#366)
b411c4d
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@backyes backyes

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
5 participants
@backyes @wangkuiyi @linrongyi @luotao1 @emailweixu