[0-size Tensor Job2 No.60] Add 0-size Tensor support for paddle.nn.functional.instance_norm [fluid_ops] #73854
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
co63oc changed the title 
Codecov ReportAttention: Patch coverage is
❌ Your patch status has failed because the patch coverage (44.82%) is below the target coverage (90.00%). You can increase the patch coverage or adjust the target coverage. Additional details and impacted files@@ Coverage Diff @@ ## develop #73854 +/- ## ========================================== Coverage ? 44.82% ========================================== Files ? 4 Lines ? 29 Branches ? 0 ========================================== Hits ? 13 Misses ? 16 Partials ? 0 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
co63oc force-pushed the g41 branch 2 times, most recently from bf3b13d to 35db413 Compare co63oc changed the title 
| @DanielSun11 CI已完成需要review |
| /re-run all-failed |
| 代码改动较大，请 @wanghuancoder review一下 |
Comment on lines 1820 to 1830
| args : (Tensor x, Tensor scale, Tensor bias, Tensor saved_mean, Tensor saved_variance, Tensor y_grad, float epsilon=1e-5) | ||
| output : Tensor(x_grad), Tensor(scale_grad), Tensor(bias_grad) | ||
| infer_meta : | ||
| func : InstanceNormGradInferMeta | ||
| spmd_rule : InstanceNormGradInferSpmd | ||
| kernel : | ||
| func : instance_norm_grad | ||
| data_type : x | ||
| optional : scale | ||
| optional : scale, bias | ||
| backward : instance_norm_double_grad | ||
| composite: instance_norm_grad(x, scale, saved_mean, saved_variance, y_grad, epsilon, x_grad, scale_grad, bias_grad) | ||
| composite: instance_norm_grad(x, scale, bias, saved_mean, saved_variance, y_grad, epsilon, x_grad, scale_grad, bias_grad) |
There was a problem hiding this comment.
这里有几个问题需要云飞评估：
- 修改反向签名是否涉及不兼容升级？
- 如果只是InferMeta用，好像有办法不修改函数签名，在kernel->func中标记需要传给Kernel的参数有哪些，可以参考lp_pool2d_grad的param。也需要云飞确认，这种用法用的比较少
- 这个bias如果只是InferMeta用，又是optional的是否可以标记为no need buffer？也需要云飞确认
There was a problem hiding this comment.
2. lp_pool2d_grad
和lp_pool2d_grad不同的是，这里同时使用了composite，使用func传参数，composite中会报错
There was a problem hiding this comment.
改反向一般不影响推理兼容性
配置no need buffer这里不太确定有没有问题，可能需要实际测试下
There was a problem hiding this comment.
已修改 增加 no_need_buffer : bias
| template <typename T> | ||
| void instance_norm_grad(const Tensor& x, | ||
| const paddle::optional<Tensor>& scale, | ||
| const paddle::optional<Tensor>& bias UNUSED, |
There was a problem hiding this comment.
不增加bias参数，composite这部分报错
zyfncg previously approved these changes Jul 21, 2025
Comment on lines 1820 to 1830
| args : (Tensor x, Tensor scale, Tensor bias, Tensor saved_mean, Tensor saved_variance, Tensor y_grad, float epsilon=1e-5) | ||
| output : Tensor(x_grad), Tensor(scale_grad), Tensor(bias_grad) | ||
| infer_meta : | ||
| func : InstanceNormGradInferMeta | ||
| spmd_rule : InstanceNormGradInferSpmd | ||
| kernel : | ||
| func : instance_norm_grad | ||
| data_type : x | ||
| optional : scale | ||
| optional : scale, bias | ||
| backward : instance_norm_double_grad | ||
| composite: instance_norm_grad(x, scale, saved_mean, saved_variance, y_grad, epsilon, x_grad, scale_grad, bias_grad) | ||
| composite: instance_norm_grad(x, scale, bias, saved_mean, saved_variance, y_grad, epsilon, x_grad, scale_grad, bias_grad) |
There was a problem hiding this comment.
改反向一般不影响推理兼容性
配置no need buffer这里不太确定有没有问题，可能需要实际测试下
| /re-run all-failed |
1 similar comment
| /re-run all-failed |
| @zyfncg 已增加no_need_buffer，CI已完成 |
zyfncg approved these changes Jul 22, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit. This suggestion is invalid because no changes were made to the code. Suggestions cannot be applied while the pull request is closed. Suggestions cannot be applied while viewing a subset of changes. Only one suggestion per line can be applied in a batch. Add this suggestion to a batch that can be applied as a single commit. Applying suggestions on deleted lines is not supported. You must change the existing code in this line in order to create a valid suggestion. Outdated suggestions cannot be applied. This suggestion has been applied or marked resolved. Suggestions cannot be applied from pending reviews. Suggestions cannot be applied on multi-line comments. Suggestions cannot be applied while the pull request is queued to merge. Suggestion cannot be applied right now. Please check back later.
PR Category
Operator Mechanism
PR Types
Bug fixes
Description
[0-size Tensor Job2 No.60] Add 0-size Tensor support for paddle.nn.functional.instance_norm [fluid_ops]
InstanceNormInferMeta 取消维度为0判断
符号推导没有对应代码
Paddle/paddle/fluid/pir/dialect/operator/interface/infer_symbolic_shape/multiary_infer_sym.cc
Line 2729 in 1b1ad5a
InstanceNormGradInferMeta 修改scale、bias中维度设置为0部分，设置为scale、bias对应维度，增加bias参数
修改 backward.yaml配置文件，kernel增加bias参数, 使用到 composite 配置，修改paddle/fluid/prim/api/composite_backward/composite_backward_api.h文件增加bias参数
修改kernel前向，反向，填充为0
增加单测
torch 测试反向填充
PaddleAPITest测试通过
