ROCm: enable cpp_extensions.load/load_inline

[t-vi]

This enables cpp_extensions.load/load_inline. This works by hipify-ing cuda sources.
Also enable tests.
CuDNN/MIOpen extensions aren't yet supported, I propose to not do this in this PR.

[dr-ci]

💊 CircleCI build failures summary and remediations

As of commit 67c784d (more details on the Dr. CI page):

---

💚 💚 Looks good so far! There are no CircleCI failures yet. 💚 💚

---

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 on the GitHub issue tracker.

See how this bot performed.

This comment has been revised 32 times.

[t-vi]

From what I can see, the errors seem to be mostly network connectivity rather than what happens in the relevant tests.

[ezyang]

some minor bugs

[t-vi]

Thank you for the prompt review, @ezyang !
I changed to keep the intermediates, that also gets rid of some of the other uglyness.
Fixed the other bits.

[t-vi]

So if you're OK with keeping the generated hip (after a bit more thought, I think it is preferable), this should have the things fixed.

[t-vi]

It the commentary OK now?

[ezyang]

Lint.

  {
    path: 'torch/utils/hipify/hipify_python.py',
    start_line: 823,
    end_line: 823,
    start_column: 17,
    end_column: 17,
    annotation_level: 'failure',
    message: '[B006] Do not use mutable data structures for argument defaults.  They are created during function definition time. All calls to the function reuse this one instance of that data structure, persisting changes between them.'
  }

[t-vi]

Oh. Sorry about that. 🤦‍♂️

[facebook-github-bot]

@ezyang is landing this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[facebook-github-bot]

@ezyang merged this pull request in d070c0b.