fix(references/depth): resize() takes interpolation=, not mode=

[Anai-Guo]

Problem

references/depth/stereo/train.py builds an eval preprocessing closure when --weights is passed, and that closure calls torchvision.transforms.functional.resize with a mode= keyword:

https://github.com/pytorch/vision/blob/main/references/depth/stereo/train.py#L225-L229

if disp is not None and not isinstance(disp, torch.Tensor):
    disp = torch.from_numpy(disp)
    if W_t != W_o:
        disp = resize(disp, (H_t, W_t), mode=InterpolationMode.BILINEAR) * scale_factor
if valid_disp_mask is not None and not isinstance(valid_disp_mask, torch.Tensor):
    valid_disp_mask = torch.from_numpy(valid_disp_mask)
    if W_t != W_o:
        valid_disp_mask = resize(valid_disp_mask, (H_t, W_t), mode=InterpolationMode.NEAREST)

resize has no mode parameter — the interpolation argument has always been called interpolation:

def resize(
    img: Tensor,
    size: list[int],
    interpolation: InterpolationMode = InterpolationMode.BILINEAR,
    max_size: Optional[int] = None,
    antialias: Optional[bool] = True,
) -> Tensor:

So both calls raise TypeError rather than resizing. The branch is guarded by W_t != W_o and by the disparity/mask still being numpy, which is why it has stayed unnoticed: it only fires when the loaded weights' transform actually rescales the width.

Reproduction

import numpy as np, torch
from torchvision.transforms.functional import InterpolationMode, resize

disp = torch.from_numpy(np.zeros((1, 10, 10), dtype="float32"))
resize(disp, (5, 5), mode=InterpolationMode.BILINEAR)

TypeError: resize() got an unexpected keyword argument 'mode'

Fix

Rename the keyword to interpolation at both call sites. Nothing else changes — the intended interpolation modes (BILINEAR for the disparity, NEAREST for the validity mask) are preserved.

Same snippet after the change:

disp = resize(disp, (5, 5), interpolation=InterpolationMode.BILINEAR)          # -> torch.Size([1, 5, 5])
resize(mask, (5, 5), interpolation=InterpolationMode.NEAREST)                  # -> torch.Size([1, 5, 5])

Both verified against an installed torchvision build, before and after.

🤖 Generated with Claude Code

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/9644

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.