deeponetTrainSteps

Training steps for training DeepONets.

trainStep(eqns, clps, bcs, usensors, network, boundary, constraint)

Main training step for purely spatial problems.

Parameters:
  • eqns (list) –

    List of equations deeponet trains to solve.

  • clps (tensor) –

    Sampled collocation points which have been processed for training.

  • bcs (tensor) –

    Sampled boundary points which have been processed for training.

  • usensors (tensor) –

    Sampled sensors which have been processed for training.

  • network (model) –

    TensorFlow deeponet to train.

  • boundary (boundaries) –

    Boundary in which model is trained over.

  • constraint (string) –

    Soft or Hard constraining network.

Returns Collocation point loss, boundary point loss, and gradients to optimize

trainStepTime(eqns, clps, bcs, ics, usensors, network, boundary, t_orders, constraint)

Main training step for purely spatial problems.

Parameters:
  • eqns (list) –

    List of equations deeponet trains to solve.

  • clps (tensor) –

    Sampled collocation points which have been processed for training.

  • bcs (tensor) –

    Sampled boundary points which have been processed for training.

  • ics (tensor) –

    Sampled initial points which have been processed for training.

  • usensors (tensor) –

    Sampled sensors which have been processed for training.

  • network (model) –

    TensorFlow deeponet to train.

  • boundary (boundaries) –

    Boundary in which model is trained over.

  • t_orders (list) –

    List of orders of time of each equation training on.

  • constraint (string) –

    Soft or Hard constraining network.

Returns Collocation point loss, boundary point loss, initial point loss, and gradients to optimize