Fix backpropagation order in optimization tutorial

[patrocinio]

Reorder optimizer.zero_grad(), loss.backward(), and optimizer.step() to match the recommended best practice documented in the tutorial.

Fixes #3507

Description

Checklist

• The issue that is being fixed is referred in the description (see above "Fixes #ISSUE_NUMBER")
• Only one issue is addressed in this pull request
• Labels from the issue that this PR is fixing are added to this pull request
• No unnecessary issues are included into this pull request.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/tutorials/3676

• 📄 Preview Python docs built from this PR

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

✅ No Failures

As of commit 88abcce with merge base ccac77f ():
💚 Looks good so far! There are no failures yet. 💚

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

[github-actions]

Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as stale.
Feel free to remove the stale label if you feel this was a mistake.
If you are unable to remove the stale label please contact a maintainer in order to do so.
If you want the bot to never mark this PR stale again, add the no-stale label.
stale pull requests will automatically be closed after 30 days of inactivity.

[sekyondaMeta]

@claude

[claude]

Claude finished @sekyondaMeta's task in 49s —— View job

---

Review of PR #3676

• Gather context and read changed files
• Review the diff against origin/main
• Check the referenced issue Feedback about Optimizing Model Parameters Page #3507
• Post review feedback

---

Summary

This PR makes a one-line change to reorder the training loop steps in beginner_source/basics/optimization_tutorial.py, moving optimizer.zero_grad() from after optimizer.step() to before loss.backward().

Before:

# Backpropagation
loss.backward()
optimizer.step()
optimizer.zero_grad()

After:

# Backpropagation
optimizer.zero_grad()
loss.backward()
optimizer.step()

Review

The change is correct and consistent with the tutorial's own prose. Lines 136-139 of the file explicitly document the three steps in this exact order:

1. Call optimizer.zero_grad() to reset gradients
2. Backpropagate with loss.backward()
3. Call optimizer.step() to adjust parameters

The code prior to this PR had the order backward() → step() → zero_grad(), which contradicts the tutorial's own explanation above it. While both orderings are functionally equivalent for a standard training loop (since zero_grad() just needs to happen before the next backward() call), having the code match the documented best practice is important in a beginner tutorial where readers are learning the canonical pattern.

No functional issues. The reordering does not change the behavior of the training loop — gradients are still zeroed before each backward() pass, and step() still runs after gradients are computed.

CI status: Lintrunner and link check pass. Build workflows are still in progress.

Verdict: This is a clean, minimal fix. LGTM.

---