[perf] feat: add optional role selection in discrete mode for NPU Profiler

[tongtong0613]

What does this PR do?

Currently, whether in end-to-end mode or discrete mode, all roles are fully collected. As the sequence length continues to increase, the volume of collected data becomes large, leading to slow parsing. Therefore, we introduce a new feature in the NPU Profiler that allows optional role selection in discrete mode, enabling quick collection of specific roles.

We have added a new roles parameter in npu_profile.yaml to specify the roles to be collected. The currently supported options are: all, rollout_generate, actor_compute_log_prob, actor_update and ref_compute_log_prob. Setting roles to ["all"] means all roles will be collected. Other options can be freely combined, for example: ["actor_update", "ref_compute_log_prob"]

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 introduces a feature for the NPU profiler, allowing selective profiling of roles in discrete mode. A critical issue exists in verl/utils/profiler/mstx_profile.py related to exception handling, where the profiler cleanup code is skipped if the decorated function throws an exception. A refactoring using a try...finally block is suggested to address this.

[gemini-code-assist]

The current implementation doesn't handle exceptions raised by the decorated function func. If func(*args, **kwargs) raises an exception, the cleanup code (mark_end_range, profile_npu.stop()) won't be executed, potentially leading to corrupted profiling data or resource leaks. Refactor the wrapper function to use a try...finally block to guarantee resource cleanup and simplify the control flow.

                profile_name = message or func.__name__
                profile_enable = self.profiler.this_step and self.profile_option is not None

                if not profile_enable:
                    return func(self, *args, **kwargs)

                discrete_mode = self.profiler.discrete
                profile_this_role = True
                # In discrete mode, check if the current role is selected for profiling.
                if discrete_mode and role is not None:
                    target_roles = self.profile_option.get("roles", [])
                    profile_this_role = "all" in target_roles or role in target_roles
                    

                # Setup profiler
                profile_npu = None
                mark_range = None
                if discrete_mode and profile_this_role:
                    profile_npu = get_npu_profiler(option=self.profile_option, role=role)
                    profile_npu.start()
                
                if profile_enable:
                    mark_range = mark_start_range(message=profile_name)

                try:
                    result = func(self, *args, **kwargs)
                finally:
                    # Teardown profiler
                    if profile_enable:
                        mark_end_range(mark_range)
                    if profile_npu:
                        profile_npu.step()
                        profile_npu.stop()
                
                return result