API Reference

Detailed reference for the unilab Python package. Every public class, function, and submodule is auto-extracted from the source tree via sphinx.ext.autodoc + autosummary. If a symbol you expect is missing, please open an issue.

Note

Reading this in a preview build that says “no API reference”? The build environment couldn’t import unilab. Install UniLab in the same venv, or read this section on the live site.

Core foundations

🧱 unilab.base

The contracts everything else depends on: TorchEnv, SimBackend, Registry, and Scene.

unilab.base — Core Contracts
🧪 unilab.envs

Manager-Based environment runtime and task-agnostic MDP terms layered on top of base.

unilab.envs — Environment runtime
🤖 unilab.tasks

Concrete locomotion, manipulation, and motion-tracking task packages.

unilab.tasks — Concrete tasks

Learning stack

🎛 Learning algorithms — unilab.rl

PPO / APPO / SAC variants in the in-tree unilab.rl stack.

Learning Algorithms — unilab.rl.algos
🏋 unilab.training

Runtime helpers, reward bookkeeping, runner orchestration.

unilab.training — Training Runtime
🔗 Shared-memory runtime — unilab.rl.runtime

Shared-memory rollout and replay primitives in unilab.rl.runtime.

Shared-Memory Runtime — unilab.rl.runtime
🧮 unilab.backend

MuJoCo and Motrix adapters that implement SimBackend.

unisim-core — Simulation Backends

Subsystems

🏞 unilab.terrains

Procedural and heightfield terrain generators.

unilab.terrains — Procedural Terrain
👁 unilab.visualization

Scene rendering and viser bridges.

unilab.visualization — Visualization
🧰 unilab.utils

Math, IO, and numerical helpers.

unilab.utils — Utilities
📝 Training logging — unilab.rl.logging

W&B / TensorBoard bridges and structured logging in unilab.rl.logging.

Training Logging — unilab.rl.logging

Top-level package