OMMX Python SDK 2.8.x

OMMX Python SDK 2.8.x#

Static Badge

See the GitHub Releases above for full details. The following summarizes the main changes.

Bug Fixes#

Correct QPLIB quadratic coefficients and dataset Artifacts (2.8.0, #1209)#

Instance.load_qplib() and ommx.qplib.load_file() now apply the format’s factor of 1/2 to every quadratic coefficient in objectives and constraints, including diagonal and cross terms. Previously, quadratic contributions were doubled. Linear terms, constants, and bounds retain their source values.

ommx.dataset.qplib() now selects ghcr.io/jij-inc/ommx/v2.8/qplib:{tag}, regenerated from the official source files with the corrected parser. The Rust loader uses the same distribution. Cached Artifacts from the old, unversioned repository do not override the new distribution. Manually saved instances must be imported again from their original .qplib files to receive the correction.

Published references are immutable, and patch releases retain their adopted distribution. MIPLIB continues to use the v2.7 distribution. See the QPLIB tutorial for the download path.

This correction changes the mathematical model, including objective values and feasibility. Recompute results obtained from affected QPLIB instances after upgrading. The 2.8.0 solver adapters require OMMX 2.8.0 or later in the v2 line.