2.8.1

A bug fix release for parsing ISO datetimes in Python < 3.11

2.8.0

A release that includes updates for providers using BackendV1.

This also includes several under the hood updates that were for v3, but are included here since this release was not expected.

v2.7.0

This release moves M3 to use the IBM Runtime SamplerV2 for executing the calibration circuits. This will break usage on other HW platforms, e.g. IQM, but I am unsure how to better handle this at present, and will push that to v3.0. There is no major feature update here save for this, so (non-IBM) users can stick with v2.6.3 if desired.

This will be the last release before a major rework resulting in M3 v3.0

Other features:

• Users can specify the execution mode used to execute the calibration circuits, i.e. users can use the same Batch or Session to run the calibration circuits and other circuits. An example is added to the documentation.

v2.6.3

This release fixes a compatibility issue with Scipy < 1.11.0 and Python 3.8.

2.6.1

A release to fix pypi uploading

2.6.0

Contains updates for Qiskit v1.0.

Breaking Changes

• Had to drop support for Qiskit Direct due to the removal of the execute function.

Fixes

• Adds noexcept tags to Cython functions returning void needed for Cython 3+

• Adds support for Python 3.12

2.5.1

M3 2.5.1

Fixes

• Relax stochastic tests (#165) for more robust deployment.

2.5

M3 2.5

Updates

• Faulty qubits now gives a warning rather than raises (#162)

Fixes

• Returning details for multiple circuits passed to apply_correction now works (#161)

2.4

M3 2.4

Features

• Splits calibrations into multiple jobs if the number of circuits is above the max supported by the device
• Ignore inoperable qubits reported by the target device.

Updates

• The total number of balanced calibration shots is increased slightly (by at most 1 per circuit) so that the precision requested by the shots argument is always faithfully met.

2.3

M3 2.3

Enhancements

• No longer raise M3Error for faulty qubits that are not being used in the mitigation itself (#153).
• Speed up column normalization calculation when doing the iterative method and a small Hamming distance correction (#154).