Install Stochastic Tools Module with libtorch support

To be able to use the stochastic tools module with libtorch, one needs to compile with the proper dependencies. For instructions see the installation guide on the website.

Testing the installation

Basic artificial neural network trainer and surrogate classes have been implemented in the stochastic tools module for testing purposes. For their documentation, see LibtorchANNTrainer and LibtorchANNSurrogate. If MOOSE compiled without error messages, we can navigate to the testing folder within the stochastic tools root and use the following command to test the implementation (in modules/stochastic_tools):

./run_tests -j 8 --re libtorch_nn

The expected output should look like this:

test:surrogates/libtorch_nn.train_and_evaluate ............................................................ OK
test:surrogates/libtorch_nn.train ......................................................................... OK
test:surrogates/libtorch_nn.evaluate ...................................................................... OK
test:surrogates/libtorch_nn.retrain ....................................................................... OK
...
...