Frequently Asked Questions¶
As GEMSEO code evolves, some calling signatures and behavior may change. These changes may break the codes that use GEMSEO and require modifications of them. See Upgrading GEMSEO for more information.
Even for simple DOEs,
GEMSEO formulates an optimization problem,
so requires a MDO formulation.
or the list of
in the order passed by the user.
This means that you must specify an objective function.
The DOE won’t try to minimize it
but it will be set as an objective in the visualizations.
For more details, we invite you to read our tutorial Tutorial: How to carry out a trade-off study.
Create a simple optimization on a single discipline¶
Available options for algorithms¶
Coupling a simulation software to GEMSEO¶
We invite you to discover all the steps in this tutorial Tutorial: How to solve an MDO problem.
Extend GEMSEO features¶
JSON schemas describe the format (i.e. structure) of JSON files, in a similar way as XML schemas define the format of XML files. JSON schemas come along with validators, that check that a JSON data structure is valid against a JSON schema, this is used in GEMSEO’ Grammars.
We invite you to read our documentation: Input and output description: grammars.
Store persistent data produced by disciplines¶
Use HDF5 caches to persist the discipline output on the disk.
We invite you to read our documentation: caching.
Error when using a HDF5 cache¶
In GEMSEO 3.2.0,
the storage of the data hashes in the HDF5 cache has been fixed
and the previous cache files are no longer valid.
If you get an error like
The file cache.h5 cannot be used because it has no file format version:
see HDF5Cache.update_file_format for converting it.,
to update the format of the file and fix the data hashes.
GEMSEO fails with openturns¶
Openturns implicitly requires the library libnsl that may not be installed by default on recent linux OSes. Under CentOS for instance, install it with:
sudo yum install libnsl
Some GEMSEO tests fail under Windows without any reason¶
The user may face some issues with the last version of Windows 10, build 2004, while running the tests. The errors are located deep in either numpy or scipy, while performing some low-level linear algebra operations. The root cause of this issue is well known and comes from an incompatibility with Windows 10, build 2004 and some versions of OpenBlas. GEMSEO users shall not encounter any issue in production. Otherwise, please contact us in order to get some mitigation instructions.
Parallel execution limitations on Windows¶
When running parallel execution tasks on Windows, the features
HDF5Cache do not work properly. This is due to the way subprocesses are forked
in this architecture. The method
is recommended as an alternative.
The progress bar may show duplicated instances during the initialization of each subprocess, in some cases it may also print the conclusion of an iteration ahead of another one that was concluded first. This is a consequence of the pickling process and does not affect the computations of the scenario.
The execution of any script using parallel execution on Windows including, but not limited to,
n_processes > 1,
DiscParallelExecution, must be protected by an
if __name__ == '__main__': statement.