Files
UniversalTrain-ai/CONTRIBUTING.md
T
chenlin c32df70326
Checks / Layout, translation and Colab checks (push) Has been cancelled
Checks / Ruff (push) Has been cancelled
Checks / Framework tests (native backend, CPU) (push) Has been cancelled
Checks / Portable Rust controller (push) Has been cancelled
release: v0.1 generic robot training and application platform
2026-11-05 11:05:01 +08:00

2.3 KiB

Contributing

Run checks from the repository root with an editable install.

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -e ".[dev]"
python -m pytest tests/ -q
python -m ruff check .

Use the agent setup guide for platform-specific installation and usage for the workflow commands.

Directory layout

mjrl-lab/
├── assets/                  # robot models and source assets
├── controller/              # input vocabulary, parsing, and device access
├── deploy/                  # native app contract and legacy target builders
│   ├── convert/             # ONNX to RKNN conversion
│   ├── dds/                 # legacy wire interface
│   └── fsm/                 # legacy Rust controller
├── docs/                    # manuals and design notes
├── rl/
│   ├── mjlab/               # vendored simulator and manager library
│   ├── mjrl/
│   │   ├── backend/         # native and Warp integration
│   │   ├── controllers/     # deterministic controller models
│   │   ├── onboarding/      # inspect, build, calibrate, verify
│   │   ├── platform/        # native app assembly and runtime
│   │   └── tasks/           # reusable MDP and curriculum code
│   └── rsl_rl/              # vendored learning algorithms
├── scenes/                  # built-in world configurations
├── scripts/                 # onboard, train, play, export, deploy
├── tasks/
│   └── templates/           # locked family recipes
├── tests/                   # regression checks and synthetic fixtures
└── tools/                   # development and validation utilities

Add robot-specific inputs under assets/<robot_id>/ or reference an existing asset checkout. Use onboarding to generate tasks/<robot_id>/robot_spec.py. Model-specific tools belong beside their asset. Reusable tools belong in tools/ or the relevant mjrl package.

Tests should pin shared contracts using neutral fixtures. Exercise a native build and command response when changing model or controller logic. Run the lock test and verify both backends when changing template behavior. Follow CLAUDE.md for authorized lock revisions and translations.