This release is a pre-release and may not be stable for production use.
ommx-openjij-adapter
Provides an adapter between OMMX and OpenJij.
Usage
ommx-openjij-adapter can be installed from PyPI:
pip install ommx-openjij-adapter
OpenJij directly accepts a Binary, unconstrained minimization model through this adapter. Prepare a constrained model explicitly before sampling it:
from ommx import DecisionVariable, Instance
from ommx_openjij_adapter import (
OMMXOpenJijSAAdapter,
OpenJijPreparationConfig,
)
x = DecisionVariable.binary(0, name="x")
instance = Instance.from_components(
decision_variables=[x],
objective=x,
constraints={0: x == 1},
sense=Instance.MINIMIZE,
)
config = OpenJijPreparationConfig(
uniform_penalty_weight=2.0,
)
prepared = OMMXOpenJijSAAdapter.prepare(instance, config=config)
prepared_samples = OMMXOpenJijSAAdapter.sample(
prepared.input,
num_reads=16,
)
sample_set = prepared.evaluate_source(prepared_samples)
print(sample_set.summary)
The finite penalty weight is a field of the OpenJijPreparationConfig passed to
prepare through config=, not an OpenJij backend sampler parameter. It must be
chosen explicitly when constraints remain after exact preparation. A finite
penalty does not guarantee that every returned sample is feasible for the source
model; inspect the feasibility recorded in the decoded SampleSet.
Input class and explicit preparation
OMMXOpenJijSAAdapter.INPUT_CLASS describes the instances that the adapter
accepts directly:
- Binary decision variables
- a polynomial objective of any degree (QUBO or Binary HUBO)
- no active regular or special constraints
- minimization
OMMXOpenJijSAAdapter.check_applicability() checks whether an instance belongs
to this input class and satisfies the adapter-specific preconditions. It does
not include preparation that the adapter can perform first. Integer
log-encoding, maximization-to-minimization conversion, exact lowering of
Indicator/OneHot/SOS1 constraints, integer slack, and finite constraint
penalties are explicit preparation operations provided by check_preparation()
and prepare().
sample() and solve() keep the common adapter contract and accept an
Instance only. Explicit preparation therefore returns an
OpenJijPreparation: pass its input Instance to the adapter, then use
evaluate_source() to evaluate the resulting samples against the source
model. The preparation itself is not an Adapter input. The report's config
field records the normalized, immutable preparation settings actually used.
The remaining fields represent one of four terminal states:
| State | source_check |
preparation_failures |
input_applicability |
|---|---|---|---|
| Source rejected | outside the preparation source class | empty | None |
| Phase rejected | accepted | non-empty, with the owning operation |
None |
| Candidate rejected | accepted | empty | non-applicable report |
| Success | accepted | empty | applicable report |
source_check is structural source-class membership. Operation availability
and preparation policy are checked by the phase that owns them and appear in
preparation_failures. steps is the prefix of OpenJij-specific operations
that completed before the terminal state; it is an operation audit, not a
separate outcome or a composed mathematical guarantee.
Common preparation policy, guarantees, and automatic selection are tracked in
OMMX issue #1111. By default,
this prototype applies only the available exact operations. Discrete integer
slack approximation requires setting allow_approximate_integer_slack=True on
OpenJijPreparationConfig; setting inequality_integer_slack_max_range alone
does not opt into approximation. Finite penalties remain an explicit operation
selected through uniform_penalty_weight or penalty_weights on that Config,
and do not assert exact constrained support.
Per-constraint penalty weights use regular constraint IDs. A model containing Indicator, OneHot, or SOS1 constraints must therefore use a uniform penalty weight after their exact lowering.
If variable bounds prove an inequality infeasible, check_preparation() and
prepare() raise the core-owned ommx.InfeasibleDetected instead of reporting
an adapter limitation. ommx.adapter.InfeasibleDetected remains an alias for
the same exception object.
The maximum of 53 auxiliary bits checked for each used Integer variable is an
availability limit of OMMX's Integer-to-Binary log-encoding operation. It is neither a
property of the OpenJij adapter's input class nor an ommx.v2.Feature. The
latter is a wire-format forward-compatibility gate that tells readers which
serialized semantics they must understand; an adapter's input class and
preconditions determine its applicability to an in-memory Instance.
OMMX does not yet implement Kind::Spin. Its addition, including direct
OpenJij Spin input, is tracked separately in
OMMX issue #1082.
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
File details
Details for the file ommx_openjij_adapter-3.0.0b2.tar.gz.
File metadata
- Download URL: ommx_openjij_adapter-3.0.0b2.tar.gz
- Upload date:
- Size: 29.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? Yes
- Uploaded via:
twine/6.1.0 CPython/3.13.14
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
36e7754a391f957e442cf9902fec754252456282e2a2cb3eaef635be24510e2c
|
|
| MD5 |
5b30c13694535084a25e1ab3141275de
|
|
| BLAKE2b-256 |
5c22791de5ffd6ea8833e27865c5dfea9de478c06a84018c7c0ed6d91c196a1a
|
Provenance
The following attestation bundles were made for ommx_openjij_adapter-3.0.0b2.tar.gz:
Publisher:
release_python.yml on Jij-Inc/ommx
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
ommx_openjij_adapter-3.0.0b2.tar.gz -
Subject digest:
36e7754a391f957e442cf9902fec754252456282e2a2cb3eaef635be24510e2c - Sigstore transparency entry: 2211583783
- Sigstore integration time:
-
Permalink:
Jij-Inc/ommx@58ba4af890afbaf9bed4b2da3375f6d46e40f1c8 -
Branch / Tag:
refs/tags/python-3.0.0b2 - Owner: https://github.com/Jij-Inc
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
release_python.yml@58ba4af890afbaf9bed4b2da3375f6d46e40f1c8 -
Trigger Event:
push
-
Statement type:
