Introduction to updates in Tunny V0.12.0.
We have improved usability by enabling the execution of processes that do not depend on components from the menu as well. If the installation of Python from a component was unsuccessful, you can now perform a reinstallation from here. Please try this if you encounter an error message such as "no optuna"
Previously, only minimization problems could be solved, but from this version, you can now choose between maximization and minimization. If not specified, all objective functions will be minimized as before. The direction can be set from the Direction input of the Construct Fish Attribute component. 1 means maximization, and -1 means minimization. If the number of inputs matches the number of objective functions, the direction can be specified for each objective function. If there is only one input, the same direction is set for all objective functions. The following example shows a setting where the first and third objective functions are minimized, and the second is maximized.
The BruteForce sampler from optuna is now available. When this sampler is selected, the optimization will either execute all possible combinations or stop when the set number of trials or elapsed time is reached. If you want to execute all combinations, it is recommended to set a large value for the number of trials or other relevant parameters. The BruteForce sampling of the BruteForce sampler performs random sampling without duplication.
We have now implemented support for DesignExplorer, developed by Thornton Tomasetti. While the previously available Optuna-Dashboard excelled in data analysis, DesignExplorer is a result verification tool that facilitates easier exploration of shapes. The ability to embed image data created with Tunny's Artifact feature makes shape exploration even more accessible.
For more details about DesignExplorer, please refer to the following:
Design ExplorerThornton Tomasetti
We have added an Info output to Tunny's components, allowing you to check the optimization state directly in Grasshopper.