APIs

Walk through RLinf’s most essential API interfaces and how to use them. These key APIs are exposed to users to simplify the complex data flows of RL, allowing them to focus on higher-level abstractions without needing to worry about the underlying implementations.

This API documentation proceeds bottom-up, starting with the foundational APIs of RLinf, including:

API	What you get
Worker	A unified interface for workers and worker groups.
Placement	An introduction to RLinf’s GPU placement strategies.
Cluster	Support for distributed training via clusters.
Channel	Low-level communication primitives, including a producer-consumer queue abstraction.

After that, we introduce the upper-layer APIs used to implement different stages of RL:

API	What you get
Actor	Actor wrappers based on FSDP and Megatron.
Rollout	Rollout wrappers built on Hugging Face and SGLang.
Env	Environment wrappers for embodied intelligence scenarios.
Data	Data structures transferred between different workers.
Embodied Data	Embodied Env/Rollout data structures.
Replay Buffer	Trajectory replay buffer design and sampling.