Updates December 18, 2023

Shunsuke Teramachi

JijZept Software Team

JijZept Updates

We’re pleased to announce the release of JijZept with important updates and fixes that further refine your experience.

Updated SDKs

• Latest Version of jijzeptlab-client - jijzeptlab-client v0.6.0 jijzeptlab-client has expanded its supported data types. The following types are now supported:

Type	Additional Information
None
int
float
str
bool
list
tuple	Note: Converted to a list on the jijzeptlab_client side
ndarray	Now accepts numpy.array
dict[str, …]
dict[tuple, …]	Can now accept QUBO/HUBO formatted dictionaries
jm.SampleSet
jm.SerializableObject	Now accepts most JijModeling objects

Additionally, when attempting to retrieve results using unsupported data types, you will receive an error message like the following:

JijZeptSolvingFailedError: ('Failed to solve the problem. The following error occurred in the solving process.', "Failed to serialize the result variable `{Result Variable Name}`: The object is not serializable. type:`{Result Variable Type}`")

• Latest Version of JijModeling - jijmodeling v1.2.0
  • In this release, we've made significant strides in improving JijModeling's performance and versatility. One of the standout changes is the transformation of MIPLIB problems into instance data and problems. We've shifted from utilizing dense coefficient matrices to primarily handling instance data as jagged arrays. This shift not only optimizes data processing but also enables the loading of significantly larger instances that were previously unmanageable.
  • Another major enhancement is the revision of the Miplib class logic. This overhaul has resulted in a more efficient execution of the Miplib.load() method, significantly speeding up the process compared to previous versions.
  • Moreover, we've addressed a critical issue in the MPS loader's parsing logic. Some instances were previously failing to load correctly due to parsing errors. This update rectifies these issues, ensuring a smoother and more reliable loading process for various instances.

New Features

• A new documentation website has been released, making it easier to learn about JijZept.Documentation Page

• We have also launched a beta version of a service to make the JijZept series more accessible to everyone.Playground

• JijSolver is now available to the general public. It will evolve as the default solver provided by Jij in the future

  • Tutorials will be added in the future.

Bug Fixes

We do not have any bugfixes in this release.