[jit] make profiling take no_grad flags into account

[zdevito]

Stack from ghstack:

• [jit] make profiling take no_grad flags into account #31071 [jit] make profiling take no_grad flags into account

Previously the profiler would think Tensors would require grad, even
when the no_grad flag is enabled during execution. This makes the profiling
and guards respect the no_grad flag, which eliminates extra differentiable
graphs that appear in the backward graph (where no_grad is typically enabled).

Differential Revision: D18915468

[kostmo]

CircleCI build failures summary

As of commit 9fefea3:

• 1/1 broken upstream at merge base cc31965 (see grid view)
  • You may want to rebase on the viable/strict branch (see its recency history):
    • Since your merge base is older than viable/strict, run these commands:

      git fetch viable/strict
      git rebase viable/strict
      

• 0/1 failures introduced in this PR

Detailed failure analysis (WIP)

Here are the reasons each build failed:

1 failure recognized by patterns:

pytorch_macos_10_13_py3_test (1/1)

Step: Test (details)

Dec 10 17:23:30 s...Traceback (most recent call last):
Dec 10 17:23:30   File "test/run_test.py", line 456, in <module>
Dec 10 17:23:30     main()
Dec 10 17:23:30   File "test/run_test.py", line 449, in main
Dec 10 17:23:30     raise RuntimeError(message)
Dec 10 17:23:30 RuntimeError: test_quantized_nn_mods failed! Received signal: SIGSEGV
Dec 10 17:23:30 + cleanup
Dec 10 17:23:30 + retcode=1
Dec 10 17:23:30 + set +x

---

This comment was automatically generated by Dr. CI.
Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions on the GitHub issue tracker.

This comment has been revised 2 times.

[Krovatkin]

[facebook-github-bot]

@zdevito merged this pull request in cc8d634.