Fix prelu_backward TensorIterator split

[zasdfgbnm]

We should have

    for (auto& sub_iter : iter.with_32bit_indexing()) {
      launch_prelu_cuda_backward_share_weights_kernel(sub_iter, weight_data);
    }

But I mistakenly wrote it as

    for (auto& sub_iter : iter.with_32bit_indexing()) {
      launch_prelu_cuda_backward_share_weights_kernel(iter, weight_data);
    }

in my previous PR. Which leads to infinite recursion on it.

I found this bug when working on #34004

I also add a TORCH_INTERNAL_ASSERT_DEBUG_ONLY to test for this.

Besides, the caller is already guaranteed contiguous, so we don't need to handle no-contiguous tensors.

[dr-ci]

💊 Build failures summary and remediations

As of commit dcdbc28 (more details on the Dr. CI page):

---

✅ None of the build failures appear to be your fault 💚

---

• 1/1 tentatively recognized as flaky ❄️
  • Click here to rerun these jobs

---

❄️ 1 tentatively flaky failure

1 failure tentatively classified as flaky but reruns have not yet been triggered to confirm:

pytorch_linux_xenial_cuda10_2_cudnn7_py3_gcc7_test (1/1)

Step: "Set Up CI Environment After attach_workspace" (full log | pattern match details | 🔁 rerun) ❄️

WARNING: infoROM is corrupted at gpu 0000:00:1D.0

|   1  Tesla M60           Off  | 00000000:00:1E.0 Off |                    0 | 
| N/A   28C    P0    37W / 150W |      0MiB /  7618MiB |     98%      Default | 
+-------------------------------+----------------------+----------------------+ 
                                                                                
+-----------------------------------------------------------------------------+ 
| Processes:                                                       GPU Memory | 
|  GPU       PID   Type   Process name                             Usage      | 
|=============================================================================| 
|  No running processes found                                                 | 
+-----------------------------------------------------------------------------+ 
WARNING: infoROM is corrupted at gpu 0000:00:1D.0 

---

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 27 times.

[zasdfgbnm]

Wait, please hold on for now, seems the testing for prelu backward is missing some cases, I will add them later in this PR.

[ailzhang]

@zasdfgbnm If you rebase on top of pytorch/master the XLA failure should be gone as well. Thanks!

[zasdfgbnm]

Caller of launch_prelu_cuda_backward_share_weights_kernel already guarantees contiguous.

[zasdfgbnm]

@VitalyFedyunin This PR is now ready. This is a bug fix for my previous prelu CUDA_tensor_apply4 PR.

[facebook-github-bot]

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

[facebook-github-bot]

@VitalyFedyunin merged this pull request in 438aed6.