27 Apr 12:40

v1.23.0 Latest

Latest

GAMSPy 1.23.0 (2026-04-27)

New features

#795: Add support for RNN for TorchSequential.
#796: Add Gated Recurrent Units (GRU) formulation.
Add support for GRU for TorchSequential.
#797: Support GAMS for statements via a context manager. The syntax is as follows: with gp.For(index, start, end, step, direction): ...
#799: Implement projection and aggregation functions in gamspy.math module.

Improvements in existing functionality

#722: Reset symbol records to its default values in case symbol.setRecords(None) is called.
#796: Add dunder str method for all Neural network formulations.

Improved documentation

#798: Add a note in the documentation about the requirement of CMU serif font to generate a pdf file with model.toLatex function.

CI/CD changes

#791: Replace pip-audit with uv audit.
#801: Use uv trixie slim images to avoid pulling bigger images than necessary.

Miscellaneous internal changes

#795: Add tests for RNN in TorchSequential formulations.
#796: Add new tests for GRU.
Add tests for GRU in TorchSequential formulations.

Assets 4

09 Apr 19:21

v1.22.0

GAMSPy 1.22.0 (2026-04-09)

New features

#770: Add Recurrent Neural Networks (RNN) formulations.
#787: Support "if" control structure of GAMS via gp.If context manager. Support "Break" and "Continue" via functions in gp.Loop.

Improvements in existing functionality

#782: Import pandas and numpy lazily.

Bug fixes

#792: UniverseAlias.records call now returns all previously registered elements in the universe instead of the labels of all the data in the container.

Improved documentation

#789: Update the installation page for on-prem license server license installation.

Miscellaneous internal changes

#770: Restructured neural network tests.

Assets 4

06 Mar 10:55

v1.21.0

GAMSPy 1.21.0 (2026-03-06)

New features

#778: Support len operation on all symbols, and operations.
#786: Support "loop" control structure of GAMS via gp.Loop context manager.

Improvements in existing functionality

#773: Allow renaming the output of model.convert function.
#775: Provide load_symbols explicitly when it's known in synch_with_gams calls.
#779: Only unload the GDX library when the GAMS system directory changes.
#783: Get rid of all suffixes if the input file to 'gamspy mps2gms' is a zip file (e.g. indus89.mps.gz). The output python file will be named indus89.py instead of indus89.mps.py.
#785: Use model attributes to return the summary of the solve instead of parsing the trace file.

Bug fixes

#784: Mitigate the issue of one environment affecting another on the same machine with package managers that keep a global cache (e.g. pixi).

Assets 4

18 Feb 17:45

v1.20.0

GAMSPy 1.20.0 (2026-02-18)

Improvements in existing functionality

#765: Return FormulationResult instance for Neural Network formulations (pooling, convolution, and linear layers).
#772: Allow specifying gdx and python files as an option in 'gamspy mps2gms' command. Add '--compress' option to 'gamspy mps2gms'.

Bug fixes

#777: Fix model attribute updates for frozen models.

Improved documentation

#764: Sync tabs in embed neural networks docs.
#771: Add a FAQ entry about running model instances in MPS or LP format.

Dependencies

#774: Upgrade gamspy_base and gamsapi dependency ranges.

Miscellaneous internal changes

#768: Use pytest's tmp_path for tests instead of creating temp directories manually.

Assets 4

06 Feb 14:19

v1.19.3

GAMSPy 1.19.3 (2026-02-06)

Bug fixes

#769: Revert the commit that causes a breaking change in the behaviour of container.loadRecordsFromGdx. If a symbol in the gdx file does not already exist in the container, create it automatically.

Assets 4

29 Jan 12:26

v1.19.2

GAMSPy 1.19.2 (2026-01-29)

Bug fixes

#766: Fix the bug in literal validation where the domain set does not have any records.

Assets 4

28 Jan 14:12

v1.19.1

GAMSPy 1.19.1 (2026-01-28)

Bug fixes

#763: Reverts breaking change in loadRecordsFromGdx that was released in 1.19.0.

Assets 4

28 Jan 09:36

v1.19.0

GAMSPy 1.19.0 (2026-01-27)

Improvements in existing functionality

#763: Accept Path-like load_from argument in container.loadRecordsFromGdx function.

Bug fixes

#763: Make loadRecordsFromGdx respect .synchronize attribute of symbols.

Improved documentation

#752: Add more examples in API reference.

CI/CD changes

#757: Replace pre-commit with prek.

Miscellaneous internal changes

#725: Add a new example showing the usage of solution pools.
#751: Add AGENTS.md file to guide coding agents.
#756: Move dev, test and doc dependencies to depedency groups instead of optional dependencies.
#758: Update the validation error message that is raised in case debugging_level is not set.

Assets 4

10 Jan 10:22

v1.18.4

GAMSPy 1.18.4 (2026-01-10)

Improvements in existing functionality

#747: Remove DE, GAMSCHK, LOGMIP, and ODHCPLEX from the list of licensed solvers when running gamspy show license.

Improved documentation

#755: Rewrite model and container user guides and improve api reference files across the board.

CI/CD changes

#748: Automate releases on Github using GitHub Actions.
#754: Cache uv and python downloads in the setup phase for subsequent jobs.

Miscellaneous internal changes

#750: Add mcp12.py to the test models.

Assets 4

19 Dec 13:03

mabdullahsoyturk

1.18.3

GAMSPy 1.18.3 (2025-12-19)

New features

#712: Integrate mps2gms to gamspy cli.

Improvements in existing functionality

#730: Read the values of most model attributes from the trace file. Lazily read others if requested.
#740: Improve the alignment of constraints in the result of model.toLatex call.

Bug fixes

#742: Do not perform definition valiation on MPSGE equations since they are generated on the GAMS side.
#743: Use gp.Number instead of a Python variable in nonbinding equations of MCP models.
#745: Retrieve universe alias name from stack in case it's not provided instead of autogenerating a name.

Assets 4