Skip to content

Parallel Computing

This repository supports parallel computing through CUDA and "multi-threading" using the joblib library.

Using Multi-threading

We utilize joblib to take advantage of multiple cores of the CPU, primarily used to accelerate the process of solving ODEs with Scipy. To set or get the number of threads, use the following code:

import PKoopmanDL

# get the number of threads
n_jobs = PKoopmanDL.get_n_jobs()

# set the number of threads
PKoopmanDL.set_n_jobs(n_jobs)

# use all available cores
PKoopmanDL.set_n_jobs(-1)

Using CUDA

CUDA is employed to accelerate neural network training, supported by pytorch. We use torch.cuda to manage the GPU resources. You can use PKoopmanDL.print_device() to check the current device.