Add 8-GPU RL-Kernel linear_logp benchmark runner

[inaniloquentee]

Summary

• add an 8-GPU Qwen3-30B-A3B 12-rollout full train+rollout benchmark runner for baseline vs RL-Kernel linear_logp
• add nsys train/rollout trace controls, including vLLM CUDA profiler start/stop for rollout capture
• add the 8-GPU config matrix and detailed runbook

Validation

• bash -n scripts/benchmarks/run-qwen3-30B-A3B-8gpu-rlk-12rollout.sh scripts/run-qwen3-30B-A3B.sh
• python -m compileall -q vime/ray/rollout.py vime/backends/vllm_utils/vllm_engine.py
• git diff --check

Note: pytest is not installed in this environment, so tests/utils/test_vllm_engine.py could not be executed here.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 98a6660b-6999-4619-9e77-e10311f1b84c

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch vime-rlk-8gpu-12rollout

---

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.}