test(pt_expt): speed up AOTInductor UT compilation and fix hardcoded device

[wanghan-iapcm]

Summary

• Set fast inductor configs (max_fusion_size=8, epilogue_fusion=False, pattern_matcher=False, package_cpp_only=True, compile_wrapper_opt_level=O0) in conftest.py to reduce .pt2 compile time by ~50% in unit tests. Tests validate correctness only, so runtime performance is irrelevant.
• Replace hardcoded torch.set_default_device("cuda:9999999") in test_deep_eval.py with torch.get_default_device() save/restore, making the device workaround resilient to changes in the fake device value.

Test plan

• Existing pt_expt tests pass (.pt2 and .pte inference, spin, fparam/aparam, change-bias, freeze)

Summary by CodeRabbit

• Tests
  • Improved test infrastructure configuration for AOTInductor/Inductor behavior.
  • Enhanced device handling in test setup to better preserve and restore PyTorch's default device settings.

Note: These are internal testing improvements with no direct impact on end-user functionality.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 84af2be0-b628-4c8a-a7a7-088d66958837

📥 Commits

Reviewing files that changed from the base of the PR and between 016ea5b and 8050d47.

📒 Files selected for processing (1)

• source/tests/pt_expt/infer/test_deep_eval.py

🚧 Files skipped from review as they are similar to previous changes (1)

• source/tests/pt_expt/infer/test_deep_eval.py

---

📝 Walkthrough

Walkthrough

Test infrastructure updates: conftest.py adds AOTInductor/Inductor configuration overrides for unit tests; test_deep_eval.py modifies test setup to preserve and restore PyTorch's default device around AOTInductor compilation operations instead of hard-coding device restoration.

Changes

Cohort / File(s)	Summary
AOTInductor Configuration source/tests/pt_expt/conftest.py	Added import of torch._inductor.config and configured multiple Inductor/AOTInductor settings including reduced max_fusion_size, disabled epilogue_fusion and pattern_matcher, and AOTInductor-specific flags (package_cpp_only, compile_wrapper_opt_level).
Device Preservation in Tests source/tests/pt_expt/infer/test_deep_eval.py	Updated test setup to capture and restore PyTorch's default device via prev = torch.get_default_device() / torch.set_default_device(prev) pattern around deserialization and AOTInductor compilation, replacing hard-coded fake CUDA device restoration.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• test(pt-expt): add .pte and .pt2 tests for dp convert-backend #5384: Directly related changes to pt_expt tests for managing PyTorch default device around AOTInductor/.pt2 compilation.

Suggested reviewers

• njzjz

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 54.55% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The PR title mentions speeding up AOTInductor UT compilation and fixing hardcoded device, which aligns with the main objectives: configuring fast AOTInductor settings and addressing device handling issues.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 016ea5b809

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.35%. Comparing base (345d162) to head (8050d47).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #5394   +/-   ##
=======================================
  Coverage   80.35%   80.35%           
=======================================
  Files         819      819           
  Lines       85445    85446    +1     
  Branches     4140     4140           
=======================================
+ Hits        68662    68663    +1     
- Misses      15508    15510    +2     
+ Partials     1275     1273    -2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.