Cheat Sheet

Use this page when you already know the workflow and need the shortest path to a working command.

Install

git clone https://github.com/RLinf/RLinf.git
cd RLinf
bash requirements/install.sh embodied --model openvla --env maniskill_libero

Start Ray

Single-node runs can start Ray locally.

ray start --head

For multi-node runs, set RLINF_NODE_RANK before ray start on every node. See Multi-Node Ray Cluster Setup.

Run Training

Launch an embodied recipe by config name.

bash examples/embodiment/run_embodiment.sh maniskill_ppo_openvla_quickstart

Evaluate

Use the unified evaluation entry point for embodied benchmarks.

bash evaluations/run_eval.sh libero/libero_spatial_openpi_pi05_eval

Next Steps

• Installation — set up RLinf and optional dependencies.

• Quick Start — run the Get Started training recipe.

• Launch & Scale — run across nodes.

• Evaluation — run standalone embodied evaluation.