Fix eval compile timing in torch.compile tutorial

[williamwen42]

Fixes pytorch/pytorch#110666.

Moved placement of torch.no_grad. Added asserts for speedup values so that future regressions are detected sooner.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/tutorials/2596

• 📄 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 1e23d37 with merge base d986068 ():
💚 Looks good so far! There are no failures yet. 💚

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

[malfet]

Looks like the fix moves torch.nograd outside of torch.compile, which is OK if it's documented somewhere, that this is nto a supported way of using torch.compile (which feels wrong to me, it should just optimize given function, i.e. respect torch.nograd decorator)

[williamwen42]

Let's push the patch for the tutorial for now, we will fix the no_grad issue in the background.