torch.reciprocal: promote integer inputs to float

[soulitzer]

Fixes #49091

[dr-ci]

💊 CI failures summary and remediations

As of commit 4d9b345 (more details on the Dr. CI page):

---

• 1/2 failures possibly* introduced in this PR
  • 1/1 non-CircleCI failure(s)
• 1/2 broken upstream at merge base 6568572 on Dec 18 from 1:48am to 9:50am

---

1 job timed out:

• pytorch_linux_bionic_py3_8_gcc9_coverage_test2

---

🚧 1 fixed upstream failure:

These were probably caused by upstream breakages that were already fixed.

Please rebase on the viable/strict branch (expand for instructions)

If your commit is older than viable/strict, run these commands:

git fetch https://github.com/pytorch/pytorch viable/strict
git rebase FETCH_HEAD

Check out the recency history of this "viable master" tracking branch.

• pytorch_linux_bionic_py3_8_gcc9_coverage_test2 on Dec 18 from 1:48am to 9:50am (1a92802 - 72b00a8)
  • 🔁 rerun

---

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions to the (internal) Dr. CI Users group.

This comment has been revised 44 times.

[mruberry]

Hey @soulitzer. Overall this looks very good. I made some inline comments.

I think this PR should also update reciprocal's documentation to clarify the integer behavior and note its different from NumPy's reciprocal. This should probably be done in a note.

[soulitzer]

Reason for disabling test_reference_numerics for bfloat16:

>>> torch.tensor(-0.0009994506803359116).reciprocal()
tensor(-1000.5496)
>>> torch.tensor(-0.0009994506803359116).reciprocal().to(torch.bfloat16)
tensor(-1000., dtype=torch.bfloat16) #`torch.bfloat16` isn't able to represent `-1000.5496`.

[mruberry]

Tests failures are unrelated.

[mruberry]

Nice work, @soulitzer! It's great that reciprocal is finally being tested with zero, and we have better undefined behavior annotations. It's also nice that we're improving over NumPy's behavior by defining reciprocal for integers sensibly.

[facebook-github-bot]

@soulitzer has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[facebook-github-bot]

@soulitzer has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[facebook-github-bot]

@soulitzer merged this pull request in 5ab9593.