Skip to content

[Phi]Fix group_norm for big tensor #74172

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
Merged
merged 4 commits into from
Jul 31, 2025
Merged

[Phi]Fix group_norm for big tensor #74172

merged 4 commits into from
Jul 31, 2025

Conversation

@ggggxm
Copy link
Contributor

@ggggxm ggggxm commented Jul 22, 2025 •
edited
Loading

PR Category

Operator Mechanism

PR Types

Bug fixes

Description

NDHWC layout的groupnorm kernel修复：

  • 只将h*w相关变量改为int64_t类型，单个C H W D等仍然为int类型，整个kernel逻辑只支持在int范围内的NDHWC参数。
  • 为kernel添加check，不支持的launch configuration拦截报错。
@paddle-bot
Copy link

paddle-bot bot commented Jul 22, 2025

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.
@ggggxm
Copy link
Contributor Author

ggggxm commented Jul 28, 2025

/re-run all-failed
lshpku
lshpku reviewed Jul 28, 2025
View reviewed changes
paddle/fluid/inference/tensorrt/plugin/group_norm_op_plugin.h Outdated
std::vector<int64_t> mean_shape_;
std::vector<int64_t> variance_shape_;
GroupNormNDHWCParams<half> params_;
GroupNormNDHWCParams<half, int64_t> params_;
Copy link
Contributor

Choose a reason for hiding this comment

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

为什么只有这个用int64？其他也改成int64会怎么样，如果是CPU端用int64也不影响性能
lshpku
lshpku reviewed Jul 28, 2025
View reviewed changes
paddle/phi/kernels/gpu/group_norm_kernel.cu Outdated
Comment on lines 329 to 330
grid.x = divUp<int32_t>(params->c, params->cPerBlock);
grid.y = divUp<int64_t>(params->dhw, params->dhwPerBlock);
Copy link
Contributor

@lshpku lshpku Jul 28, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

为什么两个除法一个是int32一个是int64？这两不都是来自同一个params吗，类型应该一样？或者其实不用显式指定模板，如果两个输入的类型一样，它会自己选择的
@ggggxm
Copy link
Contributor Author

ggggxm commented Jul 30, 2025

/re-run all-failed
1 similar comment
@ggggxm
Copy link
Contributor Author

ggggxm commented Jul 30, 2025

/re-run all-failed
wanghuancoder
wanghuancoder approved these changes Jul 31, 2025
View reviewed changes
Copy link
Contributor

@wanghuancoder wanghuancoder left a comment

Choose a reason for hiding this comment

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

LGTM
@lshpku lshpku merged commit c2a5776 into PaddlePaddle:develop Jul 31, 2025
89 of 92 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

3 participants

@ggggxm @wanghuancoder @lshpku