Skip to content
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

[Inference] Predictor support pir and new executor #58452

Merged
merged 17 commits into from
Nov 3, 2023

Conversation

yuanlehome
Copy link
Contributor

@yuanlehome yuanlehome commented Oct 27, 2023

PR types

Others

PR changes

Others

Description

推理Predictor接入新IR以及新执行器一阶段工作:

前置PR: PR#57920PR#58392PR#58074

打通流程图如下:

流程图-202311061355

结论:

  1. 与旧IR旧执行器比较,旧IR新执行器具有更少的显存占用,显存占用减少10%-30%左右,但性能一般会退化10%左右,性能优化工作进行中
  2. 与旧IR新执行器比较,新IR新执行器显存占用持平,具有更优的执行性能,性能提升10%-20%左右
  3. 因此,与旧IR旧执行器比较,新IR新执行器具有更少的显存占用、更优的执行性能

TODO:

  1. pir接入pass

Others

Pcard-71500

@paddle-bot
Copy link

paddle-bot bot commented Oct 27, 2023

你的PR提交成功,感谢你对开源项目的贡献!
请关注后续CI自动化测试结果,详情请参考Paddle-CI手册
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

@paddle-bot paddle-bot bot added the contributor External developers label Oct 27, 2023
@yuanlehome yuanlehome changed the title [Inference] Support translate to pir [Inference] Predictor support pir and new executor Nov 2, 2023
zyfncg
zyfncg previously approved these changes Nov 3, 2023
zhangbo9674
zhangbo9674 previously approved these changes Nov 3, 2023
@paddle-bot paddle-bot bot removed the contributor External developers label Nov 3, 2023
@yuanlehome yuanlehome dismissed stale reviews from zhangbo9674 and zyfncg via 4f76589 November 3, 2023 06:30
zhangbo9674
zhangbo9674 previously approved these changes Nov 3, 2023
@yuanlehome yuanlehome merged commit a5b9a74 into PaddlePaddle:develop Nov 3, 2023
28 checks passed
zeroRains pushed a commit to zeroRains/Paddle that referenced this pull request Nov 8, 2023
* support translate to pir

* fix

* update

* add ut

* disable inference_op_replace_pass

* support inplace_pass

* fix

* fix

* update

* add replace_fetch_with_shadow_output_pass

* fix comment

* update

* update
danleifeng pushed a commit to danleifeng/Paddle that referenced this pull request Nov 14, 2023
* support translate to pir

* fix

* update

* add ut

* disable inference_op_replace_pass

* support inplace_pass

* fix

* fix

* update

* add replace_fetch_with_shadow_output_pass

* fix comment

* update

* update
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants