API Reference
Auto-generated from source code. For the Rust API, see the cargo doc reference.
Python Modules
| Module |
Description |
| Algorithms |
PPO, SAC, DQN, TD3, A2C, MAPPO, DreamerV3, IMPALA, offline RL (TD3+BC, IQL, CQL, BC), LLM (GRPO, DPO), HybridPPO |
| Trainers |
High-level trainers for all algorithms (PPOTrainer, SACTrainer, MAPPOTrainer, etc.) |
| Config |
TrainingConfig, PPOConfig, SACConfig, DQNConfig -- configuration dataclasses and YAML loading |
| Buffers & Primitives |
ReplayBuffer, PrioritizedReplayBuffer, OfflineDatasetBuffer, CandleCollector |
| Callbacks & Logging |
EvalCallback, CheckpointCallback, ProgressBarCallback, loggers |
| Policies |
DiscretePolicy, ContinuousPolicy, SquashedGaussianPolicy |
| Distributed |
Multi-GPU training, gRPC actor workers, elastic scaling |
| Dashboard |
MetricsCollector, TerminalDashboard, HTMLReport -- training metrics visualisation |
Rust Crates
| Crate |
Description |
rlox-core |
VecEnv, replay/prioritized/offline buffers, GAE, V-trace, KL, GRPO, async pipeline |
rlox-nn |
Backend-agnostic NN traits (ActorCritic, QFunction, etc.) |
rlox-candle |
Candle backend: CandleActorCritic, SharedPolicy, hybrid collection |