Skip to content

Fix bicubic antialias export: use cubic_coeff_a=-0.5 instead of -0.75 #2849

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

Open
Copilot wants to merge 2 commits into
base: main
Choose a base branch
from
+7 −0
Open

Fix bicubic antialias export: use cubic_coeff_a=-0.5 instead of -0.75 #2849

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d9d49b9
Initial plan
Copilot Mar 11, 2026
0bebad4
Fix cubic_coeff_a=-0.5 for bicubic antialias=True in torchlib
Copilot Mar 11, 2026
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
7 changes: 7 additions & 0 deletions onnxscript/function_libs/torch_lib/ops/nn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2329,6 +2329,7 @@ def _aten_upsample_output_size(
mode: str,
coordinate_transformation_mode: str,
antialias: int = 0,
cubic_coeff_a: float = -0.75,
) -> TReal:
batch_and_channel = op.Shape(self, end=2, start=0)
# When output_size is passed in as a list of integers, the torch.onnx
Expand All @@ -2344,6 +2345,7 @@ def _aten_upsample_output_size(
output_size,
mode=mode,
coordinate_transformation_mode=coordinate_transformation_mode,
cubic_coeff_a=cubic_coeff_a,
nearest_mode="floor",
antialias=antialias,
)
Expand All @@ -2355,6 +2357,7 @@ def _aten_upsample_scales(
mode: str,
coordinate_transformation_mode: str,
antialias: int = 0,
cubic_coeff_a: float = -0.75,
) -> TReal:
return op.Resize(
self,
Expand All @@ -2365,6 +2368,7 @@ def _aten_upsample_scales(
None,
mode=mode,
coordinate_transformation_mode=coordinate_transformation_mode,
cubic_coeff_a=cubic_coeff_a,
nearest_mode="floor",
antialias=antialias,
)
Expand Down Expand Up @@ -2404,12 +2408,15 @@ def aten__upsample_bicubic2d_aa(
# NOTE: Based on experimentation, scales_h and scales_w are always ignored in PyTorch,
# unless when align_corners is True, in which case we do not know what is going on.
coordinate_transformation_mode = _get_upsample_align_corners_mode(align_corners)
# PyTorch uses cubic_coeff_a=-0.5 (Keys interpolation, PIL-compatible) when
# antialias=True, as opposed to -0.75 (OpenCV-compatible) for the non-antialias case.
return _aten_upsample_output_size(
self,
output_size,
mode="cubic",
coordinate_transformation_mode=coordinate_transformation_mode,
antialias=1,
cubic_coeff_a=-0.5,
)


Expand Down
Loading