Add a RolloutWorkflow to the AReaL codebase with registration and tests.
npx skills add https://github.com/inclusionAI/AReaL --skill add-workflow
Please help me install this Agent Skill. Skill: add-workflow Source: https://github.com/inclusionAI/AReaL/tree/main/.claude/skills/add-workflow Command: npx skills add https://github.com/inclusionAI/AReaL --skill add-workflow
This skill provides a clear, repeatable blueprint for adding a new RolloutWorkflow to the AReaL codebase, reducing manual setup time and human error when extending orchestrations.
Create the file areal/workflow/<name>.py implementing a RolloutWorkflow, register it, and point the trainer to the new workflow. Then run the test suite to validate.
High-intent search queries and answers about installing and using this skill.
To add a new rollout workflow to AReaL, create the workflow file in areal/workflow/, implement async methods, register it via __init__.py, update trainer configuration, and add validation tests for integration.
Registering a custom RolloutWorkflow requires exposing the new class through __init__.py and pointing the trainer configuration to the workflow file, ensuring proper imports and async operation are maintained.
Testing async rollout workflows involves adding test scaffolding that validates async behavior, checks correct tensor shapes, and verifies integration with the training loop to ensure proper orchestration execution.
Extending AReaL orchestration requires updating trainer configuration to point to the new workflow file, meaning existing trainers must be adjusted to recognize and execute the newly registered RolloutWorkflow.
AReaL workflow registration fails when the workflow file lacks proper async implementation, missing __init__.py exposure, or incomplete trainer configuration updates needed to expose the new RolloutWorkflow correctly.
A RolloutWorkflow template in AReaL is a repeatable blueprint providing ready-to-use async method structures and registration steps, reducing manual setup time and human error when extending orchestrations.
inclusionAIScaffold and register AReaL dataset loaders for SFT and RL workflows.
rishikanthcDetect wiring and integration bugs in PyTorch model implementations before training runs.
1ambdaVerify and synchronize documentation artifacts with code changes.
xchang1121Implement Triton Ascend kernels in MindSpore with custom operator registration and tensor IO wiring.
