[ducha-aiki]

What does this pull request do?

This moves the tutorials CI to PyTorch 2.15 and fixes the notebook that turns kornia's scheduled tutorials job red.

PyTorch 2.15

• .github/workflows/build.yml: pytorch-version is 2.15.0 and torchvision-version is 0.29.2 in both jobs. There is no torchvision 0.30 yet; 0.29.2 is the latest release and requires torch>=2.14.0.
• requirements.txt: torch>=2.15.0 and torchvision>=0.29.2.

adaptive_discriminator_augmentation_preset.ipynb

kornia/kornia#6305 deliberately makes the default AdaptiveDiscriminatorAugmentation stack raise on non-square images, because RandomRotation90(times=(0, 3)) can turn them by 90 degrees, and a kornia test pins that behaviour. So the notebook is what needs to change:

• The panda image is cropped to its central square, and a note says the default list needs square images.
• The training images are 64 x 128, so that cell passes an explicit stack with the same augmentations, except that RandomRotation90(times=(2, 2), p=0.5) only makes half turns. A bullet explains this.

The scheduled job checks out this repo's default branch, so the fix takes effect on kornia's next scheduled run after this merges, with no kornia change.

What did you check?

torch 2.15.0+cpu, torchvision 0.29.2+cpu, python 3.12, kornia at origin/main
  all 38 notebooks executed before the notebook fix: 37 clean, ADA failed
  ADA after the fix: runs with no errors
    plots inspected: the panda crop looks right, and real_acc tracks the 0.6 target
pre-commit passes

Not verified: this workflow itself, including kornia/workflows' env action on 2.15. This PR's CI is the first run. The other 37 notebooks were not rerun after the fix, because the fix only touches this one notebook.

Which part was done by a human, if any?

None. @ducha-aiki asked to fix the red scheduled run, asked to update the tutorials to torch 2.15, and approved opening this PR; AI did the rest. No human has reviewed the diff yet.

Written by Claude (Opus 5.5) on behalf of @ducha-aiki

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB