Add work-in-progress for visualizing gradients tutorial (issue #3186)

[j-silv]

Fixes #3186

Description

Add initial draft for visualizing gradients tutorial. Link is here

This write-up starts by discussing the difference between leaf and non-leaf tensors, and the associated requires_grad and retains_grad class attributes.

It then will go through a real-world example of visualizing gradients by using the retains_grad in a more complicated neural network like ResNet (this part is a work-in-progress).

I put the tutorial in the advanced_source directory but perhaps it would be better sorted as an intermediate tutorial or a recipe. Open to suggestions.

What's written so far is how I imagined structuring the tutorial. If you have any comments about the overall flow / material let me know. Feel free to comment on the wordage and tone as well, just know that I plan on revising tutorial as this is just the first go at it.

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/3389

• 📄 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 031ba22 with merge base 06f9c4b ():
💚 Looks good so far! There are no failures yet. 💚

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

[sekyondaMeta]

Generally seems to be headed in the right direction in terms of tone and organization from my perspective.
Can you add perquisite knowledge for this.

[soulitzer]

Thanks for the working on this tutorial. Overall I'd say though that this section (prior to the actual visualizing gradients part) can be much shorter.

By the end of this tutorial, you will be able to:

Differentiate between leaf and non-leaf tensors

have a diagram from https://github.com/szagoruyko/pytorchviz, point to the leafs

Know when to use\ retain_grad vs. ``require_grad`

"use requires_grad for leaf, use retain_grad for non-leaf"