Skip to content

[Accuracy diff No.26-27] Fix accuracy diff for paddle.linalg.(eigvals, eigvalsh) API #429

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

Merged
merged 2 commits into from
Aug 8, 2025
Merged

[Accuracy diff No.26-27] Fix accuracy diff for paddle.linalg.(eigvals, eigvalsh) API #429

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
64db839
fix eigvals and eigvalsh
ooooo-create Jul 24, 2025
bb633f6
Merge branch 'main' of https://github.com/PFCCLab/PaddleAPITest into …
ooooo-create Jul 28, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 0 additions & 9 deletions report/ci_ce_cpu/error_config.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1539,15 +1539,6 @@ paddle.linalg.eigh(Tensor([4, 4],"float64"), "L", )
paddle.linalg.eigh(Tensor([4, 4],"float64"), "U", )
paddle.linalg.eigh(Tensor([5, 5],"complex64"), )
paddle.linalg.eigh(Tensor([5, 5],"float32"), )
paddle.linalg.eigvals(Tensor([128, 128],"complex128"), name="large_x", )
paddle.linalg.eigvals(Tensor([128, 128],"complex64"), name="large_x", )
paddle.linalg.eigvals(Tensor([128, 128],"float32"), name="large_x", )
paddle.linalg.eigvals(Tensor([128, 128],"float64"), name="large_x", )
paddle.linalg.eigvals(Tensor([6, 6],"complex128"), name="small_x", )
paddle.linalg.eigvals(Tensor([6, 6],"complex64"), name="small_x", )
paddle.linalg.eigvals(Tensor([6, 6],"float64"), name="small_x", )
paddle.linalg.eigvals(x=Tensor([2, 4, 4],"float64"), )
paddle.linalg.eigvals(x=Tensor([4, 4],"float64"), )
paddle.linalg.eigvalsh(Tensor([0, 5, 5],"float32"), )
paddle.linalg.eigvalsh(Tensor([2, 5, 5],"complex64"), )
paddle.linalg.eigvalsh(Tensor([2, 5, 5],"float32"), )
Copy link
Collaborator

@cangtianhuang cangtianhuang Jul 27, 2025

Choose a reason for hiding this comment

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

paddle.linalg.eigvalsh 是否也需要删除？

Copy link
Contributor Author

@ooooo-create ooooo-create Jul 27, 2025

Choose a reason for hiding this comment

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

改完不报错的都要删嘛，eigvalsh 改成 forword only 就可以跑过了

Copy link
Collaborator

@cangtianhuang cangtianhuang Jul 27, 2025

Choose a reason for hiding this comment

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

意思是 paddle.linalg.eigvalsh 仅有反向是随机的吗？

Copy link
Contributor Author

@ooooo-create ooooo-create Jul 27, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

是的是的，api 前向只输出特征值，不会输出特征向量，算子前向第二个输出是随机，和 eigh 一样

Expand Down
2 changes: 1 addition & 1 deletion tester/api_config/config_analyzer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -768,7 +768,7 @@ def get_padding_offset(bsz, max_seq_len, seq_lens_this_time):
self.numpy_tensor = numpy.random.uniform(0.1, 1.0, size=(n_observations,)).astype(self.dtype)
else:
self.numpy_tensor = numpy.random.randint(1, 11, size=(n_observations,)).astype(self.dtype)
elif api_config.api_name.endswith("eigh"):
elif api_config.api_name.endswith("eigh") or api_config.api_name.endswith("eigvalsh"):
if self.check_arg(api_config, 0, "x"):
if len(self.shape) < 2 or self.shape[-1] != self.shape[-2]:
raise ValueError("Shape must have at least 2 dimensions and last two dimensions must be equal")
Expand Down
1 change: 1 addition & 0 deletions tester/base_config.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -295,6 +295,7 @@ forward_only_apis:
- edit_distance
- eigh
- eigvals
- eigvalsh
- embedding_grad_dense
- embedding_with_eltwise_add_xpu
- embedding_with_eltwise_add_xpu
Expand Down