[Accuracy diff No.1][BIT] Fix accuracy diff for paddle.add_n API

[cangtianhuang]

PR Category

Operator Mechanism

PR Types

Improvements

Description

The AddNKernel for paddle.add_n api on CPU (location: paddle/phi/kernels/cpu/add_n_kernel.cc) did not implement type promotion, resulting in accuracy diff compared to torch.

This pr resolves the issue by utilizing MPTypeTrait and temp_out for intermediate computations.

• When types are promoted, temp_out is requested and finally casts back to the original type.
• When types are consistent, in-place operation is retained.
• Removes the old code for adding two tensors (failed to zero out in edge cases).

All paddle.add_n test cases pass in CPU mode on PaddleAPITest:

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 91.66667% with 1 line in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@9c1900c). Learn more about missing BASE report.

⚠️ Current head 6a5377e differs from pull request most recent head 66187eb

Please upload reports for the commit 66187eb to get more accurate results.

Files with missing lines	Patch %	Lines
paddle/phi/kernels/cpu/add_n_kernel.cc	91.66%	1 Missing ⚠️

Additional details and impacted files

@@ Coverage Diff @@ ## develop #73562 +/- ## ========================================== Coverage ? 91.66% ========================================== Files ? 1 Lines ? 12 Branches ? 0 ========================================== Hits ? 11 Misses ? 1 Partials ? 0 

☔ 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.

[wanghuancoder]

LGTM 所有精度、大Tensor工作均需要书豪把把关。Coverage 9 可以找田硕豁免。