The package containes templates to include different types of neural networks in Modelica. One find blocks within the model which have to be connected with the dessired in- and outputs. The user should extend the model, give all parameters and give a interface which has to be connected to the inner blocks.

Steps to include a model:

  1. Create and train a model in TensorFlow
  2. Export a trained TensorFlow model as TfLite model
  3. Extends the appropriate base class
  4. Parametrize the model (provide path, number of in- and outputs, etc.)
  5. Define the interface and connect the in- and outputs of the blocks. The arrays have the same structure as those in TensorFlow: the inputs have to be connected in the same manner as they are used in the neural network during training.

The examples PipeLocalHeatTransfer_smallNN and TF_PI_Stateful uses this approach to create a model.



Generated at 2024-07-19T18:16:32Z by OpenModelicaOpenModelica 1.23.1 using GenerateDoc.mos