Tasks

Higher-level orchestration built on top of the runner.

Basic task

• Prepares settings
• Executes run and parses outputs
• Optionally stages inputs into a node directory for batch sweeps

Example skeleton:

# using GravityToolsNext
# t = BasicTempoTask(settings)
# res = run_task(t)

Wrapper hooks

Higher-level wrappers currently rely on a small explicit extension contract for single tasks:

• task_copy_with
• task_derive_par_output
• task_stage_inputs!

Wrappers that persist per-point results additionally use save_result_jld2.

Parameter sweeps and adaptive grids

• PriorMarginalizedTempoTask is supported with the serial scheduler.
• Adaptive2DGridTask is supported for tasks implementing the wrapper hooks above.
• Distributed scheduling remains deferred.