Skip to content

remote_handler 0.7.0

Choose a tag to compare

@nunglert nunglert released this 05 Nov 13:08
· 40 commits to main since this release

[0.7.0] — 2026-09-12

Fixed

  • Standalone collection of a collect_only=True array task no longer
    crashes with AttributeError.
    ArrayScriptTask.all_calcpaths was only
    ever set as a side effect of allocate_data(), so finish_check() broke
    for any collect_only=True task built in a fresh process that skipped
    allocation — including RemoteAtomsHandler/SinglePointHandler/
    VASPHandler/ORCAHandler. all_calcpaths is now a computed property
    derived from num_array_tasks (explicit, or inferred from
    individual_ressources), so it works identically whether or not
    allocation happened.
  • MACE trainers (MaceModelTrainer/MaceFinetuningTrainer) gained a
    collect_only mode.
    Previously update() always allocated a new array
    and get_trained_model() always (re-)submitted, so a MACE training run
    detached in one process had no way to be collected from another.

Added

  • ArrayScriptTask.from_manifest() / SingleScriptTask.from_manifest()
    — reconstruct a full, collect_only=True task from the manifest written by
    detach()/write_manifest(), without re-allocating or re-submitting.
    Distinct from TaskHandle (which only rebuilds an executor+filesystem):
    this rebuilds the actual task object, so finish_check()/
    collect_results() work exactly as they did in the original process.
  • ArrayScriptTask(num_array_tasks=...) constructor argument — lets a
    collect_only=True array task be built without individual_ressources
    (which from_manifest() relies on). Passing it alongside
    individual_ressources that disagree raises TaskError instead of
    silently picking one.
  • submit_detached() on both task classes — the second half of
    detach() (start the job, persist the manifest, return at once) without
    the allocation step, for a caller that already allocated separately (e.g.
    a MACE trainer's update()).
  • docs/source/examples/example_3_standalone.py — submit a DFT array and
    a MACE training array in one run, collect both from a later, separate run.
  • AceModelTrainer / GraceModelTrainer — ensemble-array trainers
    fitting ACE (pacemaker) / GRACE (gracemaker) potentials, the same
    shape as MaceModelTrainer. New ace / grace install extras (they
    cannot coexist in one environment — both resolve to a top-level
    tensorpotential package, but from two unrelated projects; see the
    comments in pyproject.toml), plus ace_tensorpot_gpu for the
    GPU-accelerated pacemaker backend.
  • Crash-resume now reattaches to an already-submitted job instead of
    resubmitting a duplicate
    , closing the gap where a driver restart
    mid-iteration (not just between iterations) orphaned a still-running job
    on the cluster. Sampler / Selector / GroundTruthEvaluator /
    ModelTrainer gained a pending_manifest() hook (default: reattach
    unsupported, so every existing actor keeps today's behaviour);
    ActiveLearningHandler and StagedActiveLearningHandler both gained a
    _reattach_if_pending() helper that flips a pending actor to
    collect_only for one step when it finds one (duck-typed for the
    latter, so a stage actor need not subclass the actor ABCs at all — see
    StagedActiveLearningHandler's module docstring). DetachableTaskMixin
    gained submit_if_needed() / wait_if_needed() so an actor's update()
    waits correctly either way — a single immediate finish_check() when
    freshly submitted, or polling until done (or an optional max_wait) when
    reattached — and every DetachableTaskMixin actor (the MACE/ACE/GRACE/
    NeuralilIL trainers, the VASP/ORCA/EMT evaluators) gained a matching
    reattach_max_wait constructor kwarg.
  • docs/source/active_learning.rst — a dedicated page for the
    active-learning loop: the four actor interfaces, StagedActiveLearningHandler's
    stage/ring/DAG model and resume, and every shipped evaluator/trainer/
    selector with an unfoldable, signature-checked usage example (a real
    minimal config file for the trainers that need one). docs/source/examples.rst
    gained matching Examples 4–7 (ORCA, NeuralilIL, ACE, GRACE).

Changed

  • MaceModelTrainer / MaceFinetuningTrainer / AceModelTrainer /
    GraceModelTrainer now share one _EnsembleArrayTrainer base
    (was a
    MACE-only _MaceArrayTrainer), so the three ensemble-array trainers
    duplicate only their tool-specific hooks (train-runner resolution,
    artifact naming, stub scaffolding) instead of the whole submit/poll/
    collect machinery. Public signatures unaffected.
  • docs/source/concepts.rst's active-learning section is now a short
    teaser
    pointing at the new active_learning.rst, instead of the only
    place the loop was documented.

Deprecated

  • ActiveLearningHandler — raises a DeprecationWarning on
    construction and will be removed in a future release. It hardcodes the
    fixed sampler/evaluator/trainer triple with no selector stage; use
    StagedActiveLearningHandler instead (from_actors(...) is a drop-in
    replacement for this constructor).