User guide

Use this section when you have a task package and need to perform a specific operation.

• Configuration reference — fields, defaults, and the name/params contract.
• Data — datamodules, collates, masks, and multiple loaders.
• Models — Batch/ModelOutput and built-in fusion models.
• Training — Trainer.fit, mixed precision, scheduling, and checkpoints.
• Evaluation — checkpoint loading and split metrics.
• Losses, Metrics — built-in primitives and plugin interfaces.
• Optimizers and schedulers — built-ins and config.
• Callbacks — lifecycle hooks and artifacts.
• Logging and MLflow — local and experiment tracking.
• Inference — sequential and DAG pipelines.
• Sweeps — grid and Optuna trials.
• Plugins — add task components.
• CLI workflows and Troubleshooting.