tinytorch module 09 diagrams and spacing

[vedant-a-joshi]

TinyTorch Module 09

Fixed misaligned tables. Fixed calculations for some of the comparisons. Fixed formatting of cells with extra new lines at the top as well as HTML tags. Other minor fixes to make sure everything looks neat.

Tested by running pytest tests/, and the tests relevant to module 09 passed. Ran tito module test 09 and appropriate tests were passed. Note, there were some warnings about assert vs. return in the tests. Manually verified that formatting looks good by running it as a jupyter notebook and checking manually.

Revision of PR #1885

---

By submitting this PR, you agree to release your contribution under the project's license.

[profvjreddi]

Thanks for the focused diagram pass, @vedant-a-joshi — the alignment and neatness are genuinely welcome. I'm holding it for one more round rather than merging, because a few things would regress module 09 as-is:

• Functional regression: it drops the result._grad_fn = AvgPool2dBackward(...) wiring in AvgPool2d.forward that already landed from tinytorch module 09 fixes and feature #1885, which re-breaks average-pool autograd and orphans AvgPool2dBackward as dead code. Looks like an accidental self-revert from a stale branch — a re-sync on current dev should clear it.
• Number still off (the tinytorch module 09 fixes and feature #1885 one): the worked-example Final Output shows 7 at row-2 col-0 where the position just above computes 5 (should read 5 7 9).
• Param count: the SimpleCNN Linear count should stay 20,490 — TinyTorch's Linear defaults to bias=True (its own docstring examples include bias), so 20,480 is the no-bias value.

Two that are genuinely your call, not bugs — just confirm they were intentional: switching the intro example to a depthwise per-color conv (28×28×3 / 75 params), and deleting the CNN-vs-Dense '62× fewer params' comparison block.

Sorry this took so long to get back to you — thanks for your patience, and happy to re-review as soon as it's re-synced.

[vedant-a-joshi]

Hello @profvjreddi,

I have made the updates per your comments.

You were correct about the AvgPool2d.forward error (it was stale). I also fixed the typos in the parameter count due to the bias as well as the worked-out example.

The intro example that you mentioned was a typo, I meant to change it 30x30x16. I deleted the 62x fewer params comparison block since a similar comparison is being made above it with the same model structure when showcasing parameter efficiency (showed 120x).

Please let me know what you think. I appreciate it.

[Shashank-Tripathi-07]

Hi @vedant-a-joshi, I see that your PR has been in waiting for some time and I've tagged prof. vijay to take a look into it and merge it or suggest changes to help you move forward with this. Happy Engineering !

[github-actions]

Thanks @vedant-a-joshi! 🎉

I added @vedant-a-joshi to tinytorch for: code, test, doc.

The contributor tables are now handled directly by this workflow; no follow-up command is needed.