xorl 0.1.1

eXtensible Orchestration for Reinforcement Learning

xorl

eXtensible Orchestration for Reinforcement Learning — a lightweight infrastructure for RL training.

Features

• TrainConfig — dataclass-based configuration for RL hyperparameters (PPO-style: gamma, GAE, clipping, etc.)
• Trainer — base trainer with optimizer setup, checkpointing, and training loop
• Utilities — seed management, gradient norm computation, parameter counting

Installation

pip install xorl

Quick Start

from turl import TrainConfig, Trainer

config = TrainConfig(
    lr=3e-4,
    batch_size=64,
    gamma=0.99,
    clip_range=0.2,
)

License

Apache-2.0