Skip to content

Add RL-Kernel linear_logp benchmark integration#1

Draft
inaniloquentee wants to merge 5 commits into
mainfrom
vime-rlk-integration
Draft

Add RL-Kernel linear_logp benchmark integration#1
inaniloquentee wants to merge 5 commits into
mainfrom
vime-rlk-integration

Conversation

@inaniloquentee

Copy link
Copy Markdown
Collaborator

Summary

  • expose the vime RL-Kernel integration as a linear_logp-only path for the Qwen3-30B-A3B benchmark
  • route Megatron hidden states through RL-Kernel linear_logp when the runtime is compatible, with fallback accounting for unsupported paths
  • parameterize scripts/run-qwen3-30B-A3B.sh for the 8xH100 candidate run and gate RL-Kernel with VIME_RL_KERNEL=1

Scope

Testing

  • bash -n scripts/run-qwen3-30B-A3B.sh
  • git diff --check
  • pre-commit run --files vime/utils/rl_kernel.py vime/utils/arguments.py vime/backends/megatron_utils/rl_kernel.py vime/backends/megatron_utils/loss.py vime/backends/megatron_utils/model.py scripts/run-qwen3-30B-A3B.sh tests/test_rl_kernel_args.py tests/test_rl_kernel_linear_logp_integration.py tests/test_rl_kernel_logp_integration.py
  • ...venv-vime-rlk\Scripts\python.exe -m pytest tests/test_rl_kernel_args.py tests/test_rl_kernel_linear_logp_integration.py tests/test_value_temperature.py tests/test_metric_report.py -q
  • ...venv-vime-rlk\Scripts\python.exe -m pytest tests/test_rl_kernel_logp_integration.py tests/test_rl_kernel_args.py tests/test_rl_kernel_linear_logp_integration.py -q

Not run yet

  • 8xH100 baseline
  • 8xH100 candidate with VIME_RL_KERNEL=1 and VIME_RL_KERNEL_STRICT=1

Signed-off-by: inaniloquentee <3051000145@qq.com>
Signed-off-by: inaniloquentee <3051000145@qq.com>
Signed-off-by: inaniloquentee <3051000145@qq.com>
@coderabbitai

coderabbitai Bot commented Jun 27, 2026

Copy link
Copy Markdown

Important

Review skipped

Draft detected.

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: d3d80293-7c26-493c-908f-b4154549659e

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-integration

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

Comment @coderabbitai help to get the list of available commands.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant