GpuPINNTrainer<T>

Models & Types · AiDotNet.PhysicsInformed

Provides GPU-accelerated training for Physics-Informed Neural Networks.

How It Works

For Beginners: This trainer provides GPU-accelerated training methods for PINNs. It can significantly speed up training by:

  1. Processing large batches of collocation points in parallel
  2. Performing matrix operations on the GPU
  3. Using asynchronous data transfers to overlap computation

Usage:

var architecture = new NeuralNetworkArchitecture<double>(
InputType.OneDimensional, NeuralNetworkTaskType.Regression, inputSize: 2, outputSize: 1);
var myPinn = new PhysicsInformedNeuralNetwork<double>(
architecture,
new HeatEquation<double>(),
Array.Empty<IBoundaryCondition<double>>());

var trainer = new GpuPINNTrainer<double>(myPinn);
var history = trainer.Train(epochs: 10000, learningRate: 0.001);

The trainer automatically falls back to CPU if GPU is not available.

Constructors

Constructor Summary
GpuPINNTrainer(PhysicsInformedNeuralNetwork<T> pinn, GpuPINNTrainingOptions options) Initializes a new instance of the GPU PINN trainer.

Properties

Property Summary
GpuPINNTrainingOptions Options { get; } Gets the current training options.
PhysicsInformedNeuralNetwork<T> Network { get; } Gets the underlying PINN being trained.
bool IsUsingGpu { get; } Gets whether GPU is currently being used for training.

Methods

Method Summary
GpuTrainingHistory<T> Train(Tensor<T> dataInputs, Tensor<T> dataOutputs, int epochs, double learningRate, bool verbose) Trains the PINN with GPU acceleration.
PINNGpuMemoryInfo GetGpuMemoryInfo() Gets GPU memory usage statistics.
T EvaluateLoss(Tensor<T> batchInputs, Tensor<T> batchTargets) Performs a forward pass and computes loss (does not update weights).
bool TryInitializeGpu() Attempts to initialize GPU resources.
void ReleaseGpuResources() Releases GPU resources.
void UpdateOptions(GpuPINNTrainingOptions options) Updates training options.