Dakota Reference Manual  Version 6.10
Explore and Predict with Confidence
 All Pages

Write evaluation data to HDF5 for all models that belong directly to methods


This keyword is related to the topics:


Alias: none

Argument(s): none

Default: top_method


When this option is selection, evaluation data for all models that belong directly to methods will be written to HDF5. Models that belong to other models (e.g. wrapped models in a recast relationship) are not stored.

By default, when HDF5 output is enabled, Dakota writes evaluation data only for the model that belongs to the top-level method. This keyword group is used to override the default.

HDF5 output is an experimental feature, and the contents and organization of the output file is subject to change. The current organization and a brief explanation of HDF5 is provided in the Dakota HDF5 Output section of this manual.


For the example input in the parent keyword, the following model groups would be written:

  • /models/recast/RECAST_surr_RECAST_1/
  • /models/simulation/truth_m/

The recast model, in this case, is automatically generated by Dakota to handle summing the pair of objective functions. It is the model that belongs directly to the method 'opt'. See Sources of Evaluation Data for further explanation.

The following links would be added to methods and models sources groups:

  • /methods/opt/sources/RECAST_surr_RECAST_1 $ \rightarrow $ /models/recast/RECAST_surr_RECAST_1/
  • /methods/training/sources/truth_m $ \rightarrow $ /models/simulation/truth_m/

Depending on interface_selection, the following links may also be added to model sources groups:

  • /models/simulation/truth_m/sources/truth $ \rightarrow $ /interfaces/truth/truth_m/