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.

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

Added PRelu BF16/FP32 FWD/BWD kernels #33878

Merged
merged 25 commits into from
Jul 7, 2021
Merged

Added PRelu BF16/FP32 FWD/BWD kernels #33878

merged 25 commits into from
Jul 7, 2021

Conversation

jakpiase
Copy link
Contributor

@jakpiase jakpiase commented Jun 30, 2021 •
edited
Loading

PR types

New features

PR changes

OPs

Describe

Added PRelu BF16/FP32 FWD/BWD kernels

Profiling before this PR:
image
Profiling after this PR:
image

So it made ~88% speedup, measured with Intel(R) Core(TM) i9-9940X CPU @ 3.30GHz on mobilefacenet model
It's also a solution for this issue

@paddle-bot-old
Copy link

Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

@lidanqing-intel
Copy link
Contributor

lidanqing-intel commented Jun 30, 2021 •
edited
Loading

@jakpiase Please remove the check_dygraph=False, because it will be set to False by default since following function in op_test.py

image

You can also refer to this PR #33172

@jakpiase
Copy link
Contributor Author

Thank you, I'll delete that. I am glad to know that information

@jakpiase
Copy link
Contributor Author

Unfortunately disabling check_dygraph that you have mentioned is only in check_output_with_place. The code doesn't work since in check_grad_with_place check_dygraph is not disabled.

@jakpiase
Copy link
Contributor Author

Since tests will not pass without that line, I will add it back
https://xly.bce.baidu.com/paddlepaddle/paddle/newipipe/detail/3077208/job/5391455

@lidanqing-intel
Copy link
Contributor

lidanqing-intel commented Jul 1, 2021 •
edited
Loading

@jakpiase Please remove the check_dygraph=False, because it will be set to False by default since following function in op_test.py

image

You can also refer to this PR #33172

@jakpiase What about add this pattern into check_grad_with_place ?

jczaja
jczaja previously approved these changes Jul 1, 2021
View reviewed changes
Copy link
Contributor

@jczaja jczaja left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

lidanqing-intel
lidanqing-intel previously approved these changes Jul 1, 2021
View reviewed changes
Copy link
Contributor

@lidanqing-intel lidanqing-intel left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

lidanqing-intel
lidanqing-intel approved these changes Jul 7, 2021
View reviewed changes
Copy link
Contributor

@lidanqing-intel lidanqing-intel left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

juncaipeng
juncaipeng approved these changes Jul 7, 2021
View reviewed changes
Copy link
Contributor

@juncaipeng juncaipeng left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm

@juncaipeng juncaipeng merged commit 375e561 into PaddlePaddle:develop Jul 7, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@arogowie-intel arogowie-intel arogowie-intel left review comments

@lidanqing-intel lidanqing-intel lidanqing-intel approved these changes

@juncaipeng juncaipeng juncaipeng approved these changes

@jczaja jczaja jczaja left review comments

Assignees
No one assigned
Labels
BF16 Intel
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
5 participants
@jakpiase @lidanqing-intel @jczaja @arogowie-intel @juncaipeng