Skip to main content

Ctrl+K

Overview

Version History

Key Milestones

Getting Started with AReaL-lite

Running GRPO on GSM8K Dataset

Tutorial

Installation
Quickstart
OpenAI-Compatible Workflows
Evaluation
Configurations
Quickstart (Legacy)

Best Practices

Debugging Guide
Handling OOM Issues

Customization

Dataset
Rollout and Agentic RL
Training Algorithm

References

Benchmark Guide
Reproduction Guide

Algorithms

Group Relative Policy Optimization (GRPO)
REINFORCE Leave-One-Out (RLOO)
Decoupled Clip and Dynamic Sampling Policy Optimization (DAPO)
Group Relative Policy Optimization Done Right (Dr.GRPO)
Lite-PPO

Customization (Legacy)

Dataset (Legacy)
Rollout and Agentic RL (Legacy)
Training Algorithm (Legacy)

Code Walkthrough (Legacy)

Overview
Trainer
Rollout
- Generation Server and Manager
- Rollout Worker

Repository
Open issue

.md

Trainer

Trainer#

Algorithm Interface

Allocation & Parallelism

previous

Overview

next

Master Worker

By AReaL Team

© Copyright 2025.