Training

Training in UniLab is config-first. Use the package CLI for day-to-day runs and the script entrypoints when debugging the underlying Hydra composition.

CLI reference

Routes for uv run train, uv run eval, uv run demo, and low-level scripts.

CLI Reference
Hydra config

Owner YAML layout, backend selection, and safe override examples.

Hydra Config
Logs and tracking

TensorBoard, W&B, run metadata, and trace options.

Logging
Resume and checkpoints

How algo.load_run, checkpoint files, and replay commands fit together.

Resume And Checkpoints
Docker

Run UniLab inside the checked-in Linux NVIDIA image workflow.

Docker
Multi-GPU

Current off-policy multi-GPU knobs and their config boundary.

Multi-GPU

When to Drop to scripts/train_*.py

Day-to-day runs should use the unified CLI. Reach for the low-level scripts/train_*.py entrypoints only when you are:

  • debugging a specific training stack,

  • observing Hydra compose behavior directly, or

  • comparing script-level log directories or adapter behavior.