[perf] fix: fix npu profiler and add mstx UT

[tongtong0613]

What does this PR do?

• fix the parameter passing error for profile_level
• fix the error when creating npu profiler in discrete mode
• modify the execution script
• modify ascend profiling doc
• add the discrete parameter in tool_config
• add mstx_profile UT

Checklist Before Starting

• Search for similar PRs. Paste at least one query link here: ...
• Format the PR title as [{modules}] {type}: {description} (This will be checked by the CI)
  • {modules} include fsdp, megatron, sglang, vllm, rollout, trainer, ci, training_utils, recipe, hardware, deployment, ray, worker, single_controller, misc, perf, model, algo, env, tool, ckpt, doc, data
  • If this PR involves multiple modules, separate them with , like [megatron, fsdp, doc]
  • {type} is in feat, fix, refactor, chore, test
  • If this PR breaks any API (CLI arguments, config, function signature, etc.), add [BREAKING] to the beginning of the title.
  • Example: [BREAKING][fsdp, megatron] feat: dynamic batching

Test

For changes that can not be tested by CI (e.g., algorithm implementation, new model support), validate by experiment(s) and show results like training curve plots, evaluation results, etc.

API and Usage Example

Demonstrate how the API changes if any, and provide usage example(s) if possible.

# Add code snippet or script demonstrating how to use this

Design & Code Changes

Demonstrate the high-level design if this PR is complex, and list the specific changes.

Checklist Before Submitting

Important

Please check all the following items before requesting a review, otherwise the reviewer might deprioritize this PR for review.

• Read the Contribute Guide.
• Apply pre-commit checks: pre-commit install && pre-commit run --all-files --show-diff-on-failure --color=always
• Add / Update the documentation.
• Add unit or end-to-end test(s) to the CI workflow to cover all the code. If not feasible, explain why: ...
• Once your PR is ready for CI, send a message in the ci-request channel in the verl Slack workspace. (If not accessible, please try the Feishu group (飞书群).)

[gemini-code-assist]

Code Review

This pull request provides important fixes for the NPU profiler, especially correcting the profile_level parameter passing and refactoring the profiler creation logic. The addition of comprehensive unit tests for mstx_profile is a great improvement for maintainability. The documentation and configuration changes are consistent with the code modifications. I've found one potential issue regarding data overwriting in discrete profiling mode and provided a suggestion to address it.

[gemini-code-assist]

In discrete profiling mode, if the @NPUProfiler.annotate decorator is used without providing a role, multiple profiling sessions will write to the same output directory, causing data to be overwritten. This can lead to silent data loss and incorrect profiling results.

To prevent this, I suggest using the profile_name (which defaults to the function name) as a fallback for the role when creating the profiler. This ensures that each discrete profiling session has a unique output path.

Suggested change

	profile_npu = get_npu_profiler(
	contents=self.profiler.profile_contents,
	profile_level=self.profiler.profile_level,
	profile_save_path=self.profiler.profile_save_path,
	analysis=self.profiler.analysis,
	role=role,
	)
	profile_npu = get_npu_profiler(
	contents=self.profiler.profile_contents,
	profile_level=self.profiler.profile_level,
	profile_save_path=self.profiler.profile_save_path,
	analysis=self.profiler.analysis,
	role=role or profile_name,
	)

[tongtong0613]

@vermouth1992 this pr is ready for review. 2 ci failings are not related to this pr. thanks.

[FightingZhen]

LGTM, can we merge it? @vermouth1992