[ONNX] Add support for torch.bernoulli() export (#57003)

[BowenBao]

Stack from ghstack:

• Add new_ones symbolic (#59255) #59539 Add new_ones symbolic ([ONNX] Add new_ones symbolic #59255)
• Symbolic for ReLu6 (#58560) #59538 Symbolic for ReLu6 (Symbolic for ReLu6 #58560)
• [ONNX] Sum empty tensor could not be exported to ONNX successfully. (#58141) #59537 [ONNX] Sum empty tensor could not be exported to ONNX successfully. ([ONNX] Sum empty tensor could not be exported to ONNX successfully. #58141)
• [ONNX] Add support for torch.bernoulli() export (#57003) #59536 [ONNX] Add support for torch.bernoulli() export ([ONNX] Add support for torch.bernoulli() export #57003)
• fix error message for type_as (#57948) #59535 fix error message for type_as (Improve error message for type_as #57948)
• Update Autograd Export Docs (#56594) #59534 Update Autograd Export Docs ([ONNX] Update Autograd Export Docs #56594)

Support export HuggingFace - Training DeBERTa model.

Co-authored-by: hwangdeyu deyhuang@qq.com

Differential Revision: D29046609

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 2085bd2 (more details on the Dr. CI page):

---

• 1/1 failures possibly* introduced in this PR
  • 1/1 non-scanned failure(s)

---

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.

Click here to manually regenerate this comment.

[SplitInfinity]

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

[facebook-github-bot]

@SplitInfinity merged this pull request in 83450aa.