Jarvis-HEP-Sampler

Jarvis-HEP-Sampler is the plug-in SDK for user-defined samplers in the Jarvis ecosystem.

A user writes a sampler as one Python file, loads it once, and from then on uses it in a task card exactly like a built-in method. No Jarvis-HEP source code is changed. The plug-in's points go through the same Redis queue, logging and monitor as built-in samplers. The sampler-enabled HEP branch binds each scan to an exact source snapshot and checks scientific settings before resume. Project packages include the source and its full hash; fetch verifies it and prints a load command.

Jarvis sampler init MyLatinHypercube       # fixed template -> ./MyLatinHypercube.py
#   edit draw() (and optionally OPTIONS_SCHEMA)
Jarvis sampler check MyLatinHypercube.py   # optional offline check, no Redis
Jarvis sampler load MyLatinHypercube.py    # snapshot into ~/.jarvis/samplers; reloading replaces it
Jarvis run task.yaml                     # Sampling.Method: MyLatinHypercube

The Jarvis sampler commands are part of Jarvis-HEP and use this package. They are available in the sampler-enabled Jarvis-HEP development branch. The SDK also supports offline Python use (see Python API).

Scope

Jarvis-HEP-Sampler owns the plug-in format:

• the fixed template
• the PLUGIN metadata rules and method-name rule
• loading a plug-in file
• the user registry in ~/.jarvis/samplers/
• the deterministic random streams
• the output contract (shape, real, finite, inside [0, 1])
• the log facade handed to plug-ins
• the offline driver and conformance checks

Jarvis-HEP owns the runtime:

• the Jarvis sampler CLI
• task-card validation
• the run preflight
• mapping unit-cube points to physical parameters, and selection
• Sample UUIDs, Redis, Workers and the Archiver
• checkpoints and resume
• the sampler log, telemetry and monitor
• Jarvis man and project packing

This package never imports Jarvis-HEP, Redis or Portal.

Complete package installation

Version 1.0.0. Install from PyPI:

python -m pip install "Jarvis-HEP-Sampler==1.0.0"

From this repository:

python -m pip install .

Dependencies: NumPy and jsonschema. A sampler-enabled Jarvis-HEP installs this SDK with Jarvis-HEP-Sampler>=1.0.0,<2.

The end-to-end user tutorial lives in Jarvis-Books, in the TeX manual chapter Jarvis-HEP-Sampler: Implementing Your Own Sampler (TeX/Jarvis-HEP-V2 manual).

A plug-in

"""Latin hypercube design: one point per stratum in every variable."""

import numpy as np

PLUGIN = {
    "api_version": 1,
    "method": "MyLatinHypercube",
    "version": "0.1.0",
    "profile": "independent",
}

OPTIONS_SCHEMA = {  # optional
    "type": "object",
    "properties": {"centered": {"type": "boolean"}},
    "additionalProperties": False,
}


def prepare(setup):
    rng = setup.rng()
    perms = np.stack([rng.permutation(setup.point_number) for _ in range(setup.dimension)])
    return {"perms": perms, "centered": bool(setup.options.get("centered", False))}


def draw(context, rng, plan):
    strata = plan["perms"][:, context.candidate_index]
    offset = 0.5 if plan["centered"] else rng.random(context.dimension)
    return (strata + offset) / context.point_number

Sampling:
  Method: MyLatinHypercube
  Variables:
    - name: mass
      distribution: {type: Flat, parameters: {min: 100.0, max: 1000.0}}
  Bounds:
    seed: 7
    point_number: 10000
    centered: false

The full contract is in Plug-in authoring.

A complete runnable card is provided in examples/MyLatinHypercube.task.yaml. It uses Jarvis-HEP's bundled Eggbox demonstration operator:

Jarvis sampler check examples/MyLatinHypercube.py --card examples/MyLatinHypercube.task.yaml
Jarvis sampler load examples/MyLatinHypercube.py
Jarvis validate examples/MyLatinHypercube.task.yaml
Jarvis check examples/MyLatinHypercube.task.yaml
Jarvis run examples/MyLatinHypercube.task.yaml

Jarvis sampler check is the offline file check and does not require load. Jarvis check is the calculator smoke: its identity, bound source and checkpoint stay under DATABASE/test and checkpoints/<scan>/<Method>/check/, and the formal scan's science database and resume checkpoint stay in place.

Reloading the same method updates future scans. Removing the registration preserves an existing scan's bound source and its normal Jarvis run ... --resume workflow.

Python API

from jarvis_hep_sampler import SamplerRegistry, write_template
from jarvis_hep_sampler.testing import check_plugin, run_offline

path = write_template("MyLatinHypercube")                  # Jarvis sampler init
report = check_plugin(path, dimension=2, options={})     # Jarvis sampler check
for name, result in report.rows():
    print(f"{name:14} {result}")

run = run_offline(path, variable_names=["mass", "coupling"], point_number=1000, seed=7)
print(run.points.shape, run.stop_reason)

SamplerRegistry().load(path)                             # Jarvis sampler load

Documentation

• Plug-in authoring

Status

Alpha. API version 1 supports the independent execution contract; generational metadata reserves a future profile, whose running state/feedback contract remains under design. This release runs independent plug-ins; generational (feedback samplers) arrives in a later release.

License

MIT. See LICENSE. Copyright (c) 2026 Pengxuan Zhu.