🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/tutorials/3861

• 📄 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.

@claude

Claude finished @sekyondaMeta's task in 1m 2s —— View job

Review of PR #3861

• Gather context and read changed files
• Review code changes for correctness and style
• Post review feedback

Overall: This is a clean, well-scoped modernization. The changes correctly migrate from the deprecated torchvision.transforms.ToTensor to the v2 API, and the one_hot replacement is more readable than the manual scatter_ pattern. A few observations:

Looks Good

1. v2 API usage is correct — v2.Compose([v2.ToImage(), v2.ToDtype(torch.float32, scale=True)]) is the recommended replacement for ToTensor() per the torchvision docs.
2. one_hot replacement is cleaner — torch.nn.functional.one_hot(torch.tensor(y), num_classes=10).float() is more readable and idiomatic than torch.zeros(10).scatter_(0, torch.tensor(y), value=1).
3. Prose is clear and concise — the explanatory text matches the code and links to correct documentation pages.
4. Section heading updated appropriately — "ToImage() and ToDtype()" accurately describes the new content.

Minor Suggestions (non-blocking)

1. Line 38 is long (~100 chars) — Consider breaking the target_transform lambda across lines for readability, though this is a style preference and the current form is acceptable for a tutorial:

   target_transform=v2.Lambda(
       lambda y: torch.nn.functional.one_hot(torch.tensor(y), num_classes=10).float()
   )

   Same applies to line 61 where it's repeated.

2. import torch.nn.functional as F — The PR uses torch.nn.functional.one_hot inline (twice). A common pattern in PyTorch tutorials is import torch.nn.functional as F and then F.one_hot(...). This would shorten the lambda and match conventions used elsewhere in PyTorch docs. However, since this tutorial is minimal and doesn't use F for anything else, the fully qualified name is also fine.

3. Further Reading link — The "Further Reading" section at line 70 still links to the general transforms page. Consider adding or replacing with a direct link to the v2 transforms getting started guide for readers who want to explore the new API further.

Verdict

The PR is correct, well-motivated, and addresses issue #3853. The suggestions above are minor polish — this is ready to merge as-is.

@omkar-334 Generally looks good, but can you fix the nits mentioned in the review above:

Line 38 is long (~100 chars) — Consider breaking the target_transform lambda across lines for readability, though this is a style preference and the current form is acceptable for a tutorial:

target_transform=v2.Lambda(
lambda y: torch.nn.functional.one_hot(torch.tensor(y), num_classes=10).float()
)
Same applies to line 61 where it's repeated.

import torch.nn.functional as F — The PR uses torch.nn.functional.one_hot inline (twice). A common pattern in PyTorch tutorials is import torch.nn.functional as F and then F.one_hot(...). This would shorten the lambda and match conventions used elsewhere in PyTorch docs. However, since this tutorial is minimal and doesn't use F for anything else, the fully qualified name is also fine.

Further Reading link — The "Further Reading" section at line 70 still links to the general transforms page. Consider adding or replacing with a direct link to the v2 transforms getting started guide for readers who want to explore the new API further.

hi @sekyondaMeta thanks for the review. i've applied the concerned changes and updated the branch...